feat(compiler): add cache_control breakpoints for Anthropic prompt caching#38

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
KylinMountain merged 1 commit into from
May 11, 2026
Merged

feat(compiler): add cache_control breakpoints for Anthropic prompt caching #38

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
67 changes: 51 additions & 16 deletions openkb/agent/compiler.py
View file
Open in desktop
Original file line numberDiff line numberDiff line change
Expand Up@@ -6,6 +6,13 @@
Step 3: A + summary → concepts plan (create/update/related).
Step 4: Concurrent LLM calls (A cached) → generate new + rewrite updated concepts.
Step 5: Code adds cross-ref links to related concepts, updates index.

Anthropic prompt caching is enabled via ``cache_control`` markers at two
breakpoints: end of the document message (caches system + doc across all
N+M+2 calls) and end of the assistant summary message (caches the additional
summary prefix across N+M concept-generation calls). Providers that do not
support cache_control receive a normalized list-of-blocks content payload,
which LiteLLM passes through cleanly.
"""
from __future__ import annotations

Expand DownExpand Up@@ -131,6 +138,17 @@
# LLM helpers
# ---------------------------------------------------------------------------

def _cached_text(text: str) -> list[dict]:
"""Wrap a text payload into a content-block list with an Anthropic
ephemeral cache_control marker.

LiteLLM passes the marker through to Anthropic (and OpenRouter →
Anthropic). For providers that ignore cache_control, the list-of-blocks
payload remains a valid OpenAI-compatible content shape.
"""
return [{"type": "text", "text": text, "cache_control": {"type": "ephemeral"}}]


class _Spinner:
"""Animated dots spinner that runs in a background thread."""

Expand DownExpand Up@@ -168,15 +186,23 @@ def _format_usage(elapsed: float, usage) -> str:


def _fmt_messages(messages: list[dict], max_content: int = 200) -> str:
"""Format messages for debug output, truncating long content."""
"""Format messages for debug output, truncating long content.

Accepts both plain-string content and the list-of-blocks shape used by
cache_control-tagged messages (joins all text blocks for preview).
"""
parts = []
for msg in messages:
role = msg["role"]
content = msg["content"]
if len(content) > max_content:
preview = content[:max_content] + f"... ({len(content)} chars)"
raw = msg["content"]
if isinstance(raw, list):
text = "".join(b.get("text", "") for b in raw if isinstance(b, dict))
else:
preview = content
text = raw
if len(text) > max_content:
preview = text[:max_content] + f"... ({len(text)} chars)"
else:
preview = text
parts.append(f" [{role}] {preview}")
return "\n".join(parts)

Expand All@@ -199,13 +225,15 @@ def _llm_call(model: str, messages: list[dict], step_name: str, **kwargs) -> str
return content.strip()


async def _llm_call_async(model: str, messages: list[dict], step_name: str) -> str:
async def _llm_call_async(model: str, messages: list[dict], step_name: str, **kwargs) -> str:
"""Async LLM call with timing output and debug logging."""
logger.debug("LLM request [%s]:\n%s", step_name, _fmt_messages(messages))
if kwargs:
logger.debug("LLM kwargs [%s]: %s", step_name, kwargs)

t0 = time.time()

response = await litellm.acompletion(model=model, messages=messages)
response = await litellm.acompletion(model=model, messages=messages, **kwargs)
content = response.choices[0].message.content or ""

elapsed = time.time() - t0
Expand DownExpand Up@@ -587,10 +615,14 @@ async def _compile_concepts(
# --- Step 2: Get concepts plan (A cached) ---
concept_briefs = _read_concept_briefs(wiki_dir)

# Second cache breakpoint: end of the assistant summary message. Covers
# (system + doc + summary) for the plan call and every concept call.
summary_msg = {"role": "assistant", "content": _cached_text(summary)}

plan_raw = _llm_call(model, [
system_msg,
doc_msg,
{"role": "assistant", "content": summary},
summary_msg,
{"role": "user", "content": _CONCEPTS_PLAN_USER.format(
concept_briefs=concept_briefs,
)},
Expand DownExpand Up@@ -632,7 +664,7 @@ async def _gen_create(concept: dict) -> tuple[str, str, bool, str]:
raw = await _llm_call_async(model, [
system_msg,
doc_msg,
{"role": "assistant", "content": summary},
summary_msg,
{"role": "user", "content": _CONCEPT_PAGE_USER.format(
title=title, doc_name=doc_name,
update_instruction="",
Expand DownExpand Up@@ -663,7 +695,7 @@ async def _gen_update(concept: dict) -> tuple[str, str, bool, str]:
raw = await _llm_call_async(model, [
system_msg,
doc_msg,
{"role": "assistant", "content": summary},
summary_msg,
{"role": "user", "content": _CONCEPT_UPDATE_USER.format(
title=title, doc_name=doc_name,
existing_content=existing_content,
Expand DownExpand Up@@ -741,13 +773,15 @@ async def compile_short_doc(
schema_md = get_agents_md(wiki_dir)
content = source_path.read_text(encoding="utf-8")

# Base context A: system + document
# Base context A: system + document. cache_control marker on the doc
# message creates a cache breakpoint that covers (system + doc) for
# every downstream call (summary, concepts-plan, every concept page).
system_msg = {"role": "system", "content": _SYSTEM_TEMPLATE.format(
schema_md=schema_md, language=language,
)}
doc_msg = {"role": "user", "content": _SUMMARY_USER.format(
doc_msg = {"role": "user", "content": _cached_text(_SUMMARY_USER.format(
doc_name=doc_name, content=content,
)}
))}

# --- Step 1: Generate summary ---
summary_raw = _llm_call(model, [system_msg, doc_msg], "summary")
Expand DownExpand Up@@ -792,13 +826,14 @@ async def compile_long_doc(
schema_md = get_agents_md(wiki_dir)
summary_content = summary_path.read_text(encoding="utf-8")

# Base context A
# Base context A. cache_control marker on the doc message creates a
# cache breakpoint covering (system + doc) for every concept call.
system_msg = {"role": "system", "content": _SYSTEM_TEMPLATE.format(
schema_md=schema_md, language=language,
)}
doc_msg = {"role": "user", "content": _LONG_DOC_SUMMARY_USER.format(
doc_msg = {"role": "user", "content": _cached_text(_LONG_DOC_SUMMARY_USER.format(
doc_name=doc_name, doc_id=doc_id, content=summary_content,
)}
))}

# --- Step 1: Generate overview ---
overview = _llm_call(model, [system_msg, doc_msg], "overview")
Expand Down
125 changes: 125 additions & 0 deletions tests/test_compiler.py
View file
Open in desktop
Original file line numberDiff line numberDiff line change
Expand Up@@ -651,6 +651,131 @@ async def test_handles_bad_json(self, tmp_path):
assert (wiki / "summaries" / "doc.md").exists()


class TestCacheControl:
"""Verify cache_control breakpoints are emitted on the right messages
so Anthropic prompt caching can hit on every reuse of the base context.
"""

@staticmethod
def _has_cache_breakpoint(message: dict) -> bool:
content = message.get("content")
if not isinstance(content, list):
return False
return any(
isinstance(b, dict) and b.get("cache_control", {}).get("type") == "ephemeral"
for b in content
)

@pytest.mark.asyncio
async def test_short_doc_marks_doc_and_summary(self, tmp_path):
wiki = tmp_path / "wiki"
(wiki / "sources").mkdir(parents=True)
(wiki / "summaries").mkdir(parents=True)
(wiki / "concepts").mkdir(parents=True)
(wiki / "index.md").write_text(
"# Index\n\n## Documents\n\n## Concepts\n", encoding="utf-8",
)
src = wiki / "sources" / "doc.md"
src.write_text("Body text about caching.", encoding="utf-8")
(tmp_path / ".openkb").mkdir()

summary_response = json.dumps({"brief": "B", "content": "summary body"})
plan_response = json.dumps({
"create": [{"name": "topic", "title": "Topic"}],
"update": [], "related": [],
})
concept_response = json.dumps({"brief": "C", "content": "page body"})

captured_sync_calls: list[list[dict]] = []
captured_async_calls: list[list[dict]] = []

sync_responses = [summary_response, plan_response]

def sync_side_effect(*args, **kwargs):
captured_sync_calls.append(kwargs["messages"])
idx = min(len(captured_sync_calls) - 1, len(sync_responses) - 1)
mock_resp = MagicMock()
mock_resp.choices = [MagicMock()]
mock_resp.choices[0].message.content = sync_responses[idx]
mock_resp.usage = MagicMock(prompt_tokens=1, completion_tokens=1)
mock_resp.usage.prompt_tokens_details = None
return mock_resp

async def async_side_effect(*args, **kwargs):
captured_async_calls.append(kwargs["messages"])
mock_resp = MagicMock()
mock_resp.choices = [MagicMock()]
mock_resp.choices[0].message.content = concept_response
mock_resp.usage = MagicMock(prompt_tokens=1, completion_tokens=1)
mock_resp.usage.prompt_tokens_details = None
return mock_resp

with patch("openkb.agent.compiler.litellm") as mock_litellm:
mock_litellm.completion = MagicMock(side_effect=sync_side_effect)
mock_litellm.acompletion = AsyncMock(side_effect=async_side_effect)
await compile_short_doc("doc", src, tmp_path, "anthropic/claude-sonnet-4-5")

# Step 1 (summary): doc_msg carries the breakpoint.
summary_call = captured_sync_calls[0]
assert summary_call[0]["role"] == "system"
assert summary_call[1]["role"] == "user"
assert self._has_cache_breakpoint(summary_call[1]), (
"doc_msg in summary call must carry an ephemeral cache_control marker"
)

# Step 2 (plan): doc_msg AND assistant summary both carry breakpoints.
plan_call = captured_sync_calls[1]
assert self._has_cache_breakpoint(plan_call[1])
assert plan_call[2]["role"] == "assistant"
assert self._has_cache_breakpoint(plan_call[2]), (
"assistant summary in plan call must carry a cache_control marker"
)

# Step 3 (concept generation): same two breakpoints reused.
assert captured_async_calls, "expected at least one async concept call"
concept_call = captured_async_calls[0]
assert self._has_cache_breakpoint(concept_call[1])
assert self._has_cache_breakpoint(concept_call[2])

@pytest.mark.asyncio
async def test_long_doc_marks_doc_message(self, tmp_path):
wiki = tmp_path / "wiki"
(wiki / "summaries").mkdir(parents=True)
(wiki / "concepts").mkdir(parents=True)
(wiki / "index.md").write_text(
"# Index\n\n## Documents\n\n## Concepts\n", encoding="utf-8",
)
sp = wiki / "summaries" / "big.md"
sp.write_text("PageIndex tree summary.", encoding="utf-8")
(tmp_path / ".openkb").mkdir()

captured: list[list[dict]] = []
plan_response = json.dumps({"create": [], "update": [], "related": []})

def sync_side_effect(*args, **kwargs):
captured.append(kwargs["messages"])
mock_resp = MagicMock()
mock_resp.choices = [MagicMock()]
# First call: overview (plain text); second: plan (JSON).
mock_resp.choices[0].message.content = (
"Overview text" if len(captured) == 1 else plan_response
)
mock_resp.usage = MagicMock(prompt_tokens=1, completion_tokens=1)
mock_resp.usage.prompt_tokens_details = None
return mock_resp

with patch("openkb.agent.compiler.litellm") as mock_litellm:
mock_litellm.completion = MagicMock(side_effect=sync_side_effect)
mock_litellm.acompletion = AsyncMock()
await compile_long_doc(
"big", sp, "doc-id-1", tmp_path, "anthropic/claude-sonnet-4-5",
)

overview_call = captured[0]
assert overview_call[1]["role"] == "user"
assert self._has_cache_breakpoint(overview_call[1])


class TestCompileLongDoc:
@pytest.mark.asyncio
async def test_full_pipeline(self, tmp_path):
Expand Down
, 'i'); if (__m === '*' || __re.test(location.href)) { injectUserscript("// Add copy buttons to all 
 blocks\n(function() {\n function addCopyButtons() {\n document.querySelectorAll('pre code').forEach(function(codeBlock) {\n if (codeBlock.parentElement.hasAttribute('data-copy-added')) return;\n codeBlock.parentElement.setAttribute('data-copy-added', 'true');\n \n var btn = document.createElement('button');\n btn.textContent = 'Copy';\n btn.style.cssText = 'position:absolute;top:4px;right:4px;padding:2px 8px;font-size:11px;background:#4ecdc4;border:none;border-radius:4px;color:#1a1a2e;cursor:pointer;opacity:0.7;transition:opacity 0.2s;';\n btn.onmouseover = function() { this.style.opacity = '1'; };\n btn.onmouseout = function() { this.style.opacity = '0.7'; };\n btn.onclick = function() {\n navigator.clipboard.writeText(codeBlock.textContent).then(function() {\n btn.textContent = 'Copied!';\n setTimeout(function() { btn.textContent = 'Copy'; }, 1500);\n });\n };\n codeBlock.parentElement.style.position = 'relative';\n codeBlock.parentElement.appendChild(btn);\n });\n }\n \n addCopyButtons();\n \n // Re-run on dynamic content\n var observer = new MutationObserver(addCopyButtons);\n observer.observe(document.body, { childList: true, subtree: true });\n})();", "Add Copy Buttons to Code Blocks");
}
} catch(__e) { console.warn('[Userscript:Add Copy Buttons to Code Blocks]', __e); }
})();
(function(){
try {
var __m = "github.com";
var __re = new RegExp('^' + "github\\.com" + '

Skip to content
feat(compiler): add cache_control breakpoints for Anthropic prompt caching#38

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

 Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and
privacy statement. We’ll occasionally send you account related emails.

Already on GitHub?
Sign in
to your account

 Merged

KylinMountain
merged 1 commit into

from

May 11, 2026
 Merged


feat(compiler): add cache_control breakpoints for Anthropic prompt caching
#38
Show file tree
Hide file tree
 Changes from all commits
Commits

Show all changes


1 commit


Select commit


File filter

Filter by extension
Filter by extension
Conversations

Failed to load comments. 
Loading
Jump to

Jump to file


Failed to load files. 
Loading
 Diff view

 Diff view


67 changes: 51 additions & 16 deletions
openkb/agent/compiler.py

View file


Open in desktop

Original file line numberDiff line numberDiff line change
Expand Up@@ -6,6 +6,13 @@
 Step 3: A + summary → concepts plan (create/update/related).
 Step 4: Concurrent LLM calls (A cached) → generate new + rewrite updated concepts.
 Step 5: Code adds cross-ref links to related concepts, updates index.

Anthropic prompt caching is enabled via ``cache_control`` markers at two
breakpoints: end of the document message (caches system + doc across all
N+M+2 calls) and end of the assistant summary message (caches the additional
summary prefix across N+M concept-generation calls). Providers that do not
support cache_control receive a normalized list-of-blocks content payload,
which LiteLLM passes through cleanly.
"""
from __future__ import annotations

Expand DownExpand Up@@ -131,6 +138,17 @@
# LLM helpers
# ---------------------------------------------------------------------------

def _cached_text(text: str) -> list[dict]:
 """Wrap a text payload into a content-block list with an Anthropic
 ephemeral cache_control marker.

 LiteLLM passes the marker through to Anthropic (and OpenRouter →
 Anthropic). For providers that ignore cache_control, the list-of-blocks
 payload remains a valid OpenAI-compatible content shape.
 """
 return [{"type": "text", "text": text, "cache_control": {"type": "ephemeral"}}]


class _Spinner:
 """Animated dots spinner that runs in a background thread."""

Expand DownExpand Up@@ -168,15 +186,23 @@ def _format_usage(elapsed: float, usage) -> str:


def _fmt_messages(messages: list[dict], max_content: int = 200) -> str:
 """Format messages for debug output, truncating long content."""
 """Format messages for debug output, truncating long content.

 Accepts both plain-string content and the list-of-blocks shape used by
 cache_control-tagged messages (joins all text blocks for preview).
 """
 parts = []
 for msg in messages:
 role = msg["role"]
content = msg["content"]
 if len(content) > max_content:
preview = content[:max_content] + f"... ({len(content)} chars)"
raw = msg["content"]
 if isinstance(raw, list):
text = "".join(b.get("text", "") for b in raw if isinstance(b, dict))
 else:
 preview = content
 text = raw
 if len(text) > max_content:
 preview = text[:max_content] + f"... ({len(text)} chars)"
 else:
 preview = text
 parts.append(f" [{role}] {preview}")
 return "\n".join(parts)

Expand All@@ -199,13 +225,15 @@ def _llm_call(model: str, messages: list[dict], step_name: str, **kwargs) -> str
 return content.strip()


async def _llm_call_async(model: str, messages: list[dict], step_name: str) -> str:
async def _llm_call_async(model: str, messages: list[dict], step_name: str, **kwargs) -> str:
 """Async LLM call with timing output and debug logging."""
 logger.debug("LLM request [%s]:\n%s", step_name, _fmt_messages(messages))
 if kwargs:
 logger.debug("LLM kwargs [%s]: %s", step_name, kwargs)

 t0 = time.time()

 response = await litellm.acompletion(model=model, messages=messages)
 response = await litellm.acompletion(model=model, messages=messages, **kwargs)
 content = response.choices[0].message.content or ""

 elapsed = time.time() - t0
Expand DownExpand Up@@ -587,10 +615,14 @@ async def _compile_concepts(
 # --- Step 2: Get concepts plan (A cached) ---
 concept_briefs = _read_concept_briefs(wiki_dir)

 # Second cache breakpoint: end of the assistant summary message. Covers
 # (system + doc + summary) for the plan call and every concept call.
 summary_msg = {"role": "assistant", "content": _cached_text(summary)}

 plan_raw = _llm_call(model, [
 system_msg,
 doc_msg,
{"role": "assistant", "content": summary},
summary_msg,
 {"role": "user", "content": _CONCEPTS_PLAN_USER.format(
 concept_briefs=concept_briefs,
 )},
Expand DownExpand Up@@ -632,7 +664,7 @@ async def _gen_create(concept: dict) -> tuple[str, str, bool, str]:
 raw = await _llm_call_async(model, [
 system_msg,
 doc_msg,
{"role": "assistant", "content": summary},
summary_msg,
 {"role": "user", "content": _CONCEPT_PAGE_USER.format(
 title=title, doc_name=doc_name,
 update_instruction="",
Expand DownExpand Up@@ -663,7 +695,7 @@ async def _gen_update(concept: dict) -> tuple[str, str, bool, str]:
 raw = await _llm_call_async(model, [
 system_msg,
 doc_msg,
{"role": "assistant", "content": summary},
summary_msg,
 {"role": "user", "content": _CONCEPT_UPDATE_USER.format(
 title=title, doc_name=doc_name,
 existing_content=existing_content,
Expand DownExpand Up@@ -741,13 +773,15 @@ async def compile_short_doc(
 schema_md = get_agents_md(wiki_dir)
 content = source_path.read_text(encoding="utf-8")

 # Base context A: system + document
 # Base context A: system + document. cache_control marker on the doc
 # message creates a cache breakpoint that covers (system + doc) for
 # every downstream call (summary, concepts-plan, every concept page).
 system_msg = {"role": "system", "content": _SYSTEM_TEMPLATE.format(
 schema_md=schema_md, language=language,
 )}
 doc_msg = {"role": "user", "content": _SUMMARY_USER.format(
 doc_msg = {"role": "user", "content": _cached_text(_SUMMARY_USER.format(
 doc_name=doc_name, content=content,
 )}
 ))}

 # --- Step 1: Generate summary ---
 summary_raw = _llm_call(model, [system_msg, doc_msg], "summary")
Expand DownExpand Up@@ -792,13 +826,14 @@ async def compile_long_doc(
 schema_md = get_agents_md(wiki_dir)
 summary_content = summary_path.read_text(encoding="utf-8")

 # Base context A
 # Base context A. cache_control marker on the doc message creates a
 # cache breakpoint covering (system + doc) for every concept call.
 system_msg = {"role": "system", "content": _SYSTEM_TEMPLATE.format(
 schema_md=schema_md, language=language,
 )}
 doc_msg = {"role": "user", "content": _LONG_DOC_SUMMARY_USER.format(
 doc_msg = {"role": "user", "content": _cached_text(_LONG_DOC_SUMMARY_USER.format(
 doc_name=doc_name, doc_id=doc_id, content=summary_content,
 )}
 ))}

 # --- Step 1: Generate overview ---
 overview = _llm_call(model, [system_msg, doc_msg], "overview")
Expand Down

125 changes: 125 additions & 0 deletions
tests/test_compiler.py

View file


Open in desktop

Original file line numberDiff line numberDiff line change
Expand Up@@ -651,6 +651,131 @@ async def test_handles_bad_json(self, tmp_path):
 assert (wiki / "summaries" / "doc.md").exists()


class TestCacheControl:
 """Verify cache_control breakpoints are emitted on the right messages
 so Anthropic prompt caching can hit on every reuse of the base context.
 """

 @staticmethod
 def _has_cache_breakpoint(message: dict) -> bool:
 content = message.get("content")
 if not isinstance(content, list):
 return False
 return any(
 isinstance(b, dict) and b.get("cache_control", {}).get("type") == "ephemeral"
 for b in content
 )

 @pytest.mark.asyncio
 async def test_short_doc_marks_doc_and_summary(self, tmp_path):
 wiki = tmp_path / "wiki"
 (wiki / "sources").mkdir(parents=True)
 (wiki / "summaries").mkdir(parents=True)
 (wiki / "concepts").mkdir(parents=True)
 (wiki / "index.md").write_text(
 "# Index\n\n## Documents\n\n## Concepts\n", encoding="utf-8",
 )
 src = wiki / "sources" / "doc.md"
 src.write_text("Body text about caching.", encoding="utf-8")
 (tmp_path / ".openkb").mkdir()

 summary_response = json.dumps({"brief": "B", "content": "summary body"})
 plan_response = json.dumps({
 "create": [{"name": "topic", "title": "Topic"}],
 "update": [], "related": [],
 })
 concept_response = json.dumps({"brief": "C", "content": "page body"})

 captured_sync_calls: list[list[dict]] = []
 captured_async_calls: list[list[dict]] = []

 sync_responses = [summary_response, plan_response]

 def sync_side_effect(*args, **kwargs):
 captured_sync_calls.append(kwargs["messages"])
 idx = min(len(captured_sync_calls) - 1, len(sync_responses) - 1)
 mock_resp = MagicMock()
 mock_resp.choices = [MagicMock()]
 mock_resp.choices[0].message.content = sync_responses[idx]
 mock_resp.usage = MagicMock(prompt_tokens=1, completion_tokens=1)
 mock_resp.usage.prompt_tokens_details = None
 return mock_resp

 async def async_side_effect(*args, **kwargs):
 captured_async_calls.append(kwargs["messages"])
 mock_resp = MagicMock()
 mock_resp.choices = [MagicMock()]
 mock_resp.choices[0].message.content = concept_response
 mock_resp.usage = MagicMock(prompt_tokens=1, completion_tokens=1)
 mock_resp.usage.prompt_tokens_details = None
 return mock_resp

 with patch("openkb.agent.compiler.litellm") as mock_litellm:
 mock_litellm.completion = MagicMock(side_effect=sync_side_effect)
 mock_litellm.acompletion = AsyncMock(side_effect=async_side_effect)
 await compile_short_doc("doc", src, tmp_path, "anthropic/claude-sonnet-4-5")

 # Step 1 (summary): doc_msg carries the breakpoint.
 summary_call = captured_sync_calls[0]
 assert summary_call[0]["role"] == "system"
 assert summary_call[1]["role"] == "user"
 assert self._has_cache_breakpoint(summary_call[1]), (
 "doc_msg in summary call must carry an ephemeral cache_control marker"
 )

 # Step 2 (plan): doc_msg AND assistant summary both carry breakpoints.
 plan_call = captured_sync_calls[1]
 assert self._has_cache_breakpoint(plan_call[1])
 assert plan_call[2]["role"] == "assistant"
 assert self._has_cache_breakpoint(plan_call[2]), (
 "assistant summary in plan call must carry a cache_control marker"
 )

 # Step 3 (concept generation): same two breakpoints reused.
 assert captured_async_calls, "expected at least one async concept call"
 concept_call = captured_async_calls[0]
 assert self._has_cache_breakpoint(concept_call[1])
 assert self._has_cache_breakpoint(concept_call[2])

 @pytest.mark.asyncio
 async def test_long_doc_marks_doc_message(self, tmp_path):
 wiki = tmp_path / "wiki"
 (wiki / "summaries").mkdir(parents=True)
 (wiki / "concepts").mkdir(parents=True)
 (wiki / "index.md").write_text(
 "# Index\n\n## Documents\n\n## Concepts\n", encoding="utf-8",
 )
 sp = wiki / "summaries" / "big.md"
 sp.write_text("PageIndex tree summary.", encoding="utf-8")
 (tmp_path / ".openkb").mkdir()

 captured: list[list[dict]] = []
 plan_response = json.dumps({"create": [], "update": [], "related": []})

 def sync_side_effect(*args, **kwargs):
 captured.append(kwargs["messages"])
 mock_resp = MagicMock()
 mock_resp.choices = [MagicMock()]
 # First call: overview (plain text); second: plan (JSON).
 mock_resp.choices[0].message.content = (
 "Overview text" if len(captured) == 1 else plan_response
 )
 mock_resp.usage = MagicMock(prompt_tokens=1, completion_tokens=1)
 mock_resp.usage.prompt_tokens_details = None
 return mock_resp

 with patch("openkb.agent.compiler.litellm") as mock_litellm:
 mock_litellm.completion = MagicMock(side_effect=sync_side_effect)
 mock_litellm.acompletion = AsyncMock()
 await compile_long_doc(
 "big", sp, "doc-id-1", tmp_path, "anthropic/claude-sonnet-4-5",
 )

 overview_call = captured[0]
 assert overview_call[1]["role"] == "user"
 assert self._has_cache_breakpoint(overview_call[1])


class TestCompileLongDoc:
 @pytest.mark.asyncio
 async def test_full_pipeline(self, tmp_path):
Expand Down

, 'i');
if (__m === '*' || __re.test(location.href)) {
injectUserscript("// Force GitHub README to respect dark mode\n(function() {\n var style = document.createElement('style');\n style.textContent = '\n .markdown-body {\n color-scheme: dark light;\n }\n .markdown-body pre { background: #161b22 !important; }\n .markdown-body code { background: rgba(110, 118, 129, 0.4) !important; }\n .markdown-body table th, .markdown-body table td { border-color: #30363d !important; }\n .markdown-body img { background: #0d1117; }\n .markdown-body blockquote { border-left-color: #8b949e; }\n .markdown-body hr { border-color: #30363d; }\n ';\n document.head.appendChild(style);\n})();", "GitHub Dark Mode README Fix");
}
} catch(__e) { console.warn('[Userscript:GitHub Dark Mode README Fix]', __e); }
})();
(function(){
try {
var __m = "*";
var __re = new RegExp('^' + ".*" + '

Skip to content
feat(compiler): add cache_control breakpoints for Anthropic prompt caching#38

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

 Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and
privacy statement. We’ll occasionally send you account related emails.

Already on GitHub?
Sign in
to your account

 Merged

KylinMountain
merged 1 commit into

from

May 11, 2026
 Merged


feat(compiler): add cache_control breakpoints for Anthropic prompt caching
#38
Show file tree
Hide file tree
 Changes from all commits
Commits

Show all changes


1 commit


Select commit


File filter

Filter by extension
Filter by extension
Conversations

Failed to load comments. 
Loading
Jump to

Jump to file


Failed to load files. 
Loading
 Diff view

 Diff view


67 changes: 51 additions & 16 deletions
openkb/agent/compiler.py

View file


Open in desktop

Original file line numberDiff line numberDiff line change
Expand Up@@ -6,6 +6,13 @@
 Step 3: A + summary → concepts plan (create/update/related).
 Step 4: Concurrent LLM calls (A cached) → generate new + rewrite updated concepts.
 Step 5: Code adds cross-ref links to related concepts, updates index.

Anthropic prompt caching is enabled via ``cache_control`` markers at two
breakpoints: end of the document message (caches system + doc across all
N+M+2 calls) and end of the assistant summary message (caches the additional
summary prefix across N+M concept-generation calls). Providers that do not
support cache_control receive a normalized list-of-blocks content payload,
which LiteLLM passes through cleanly.
"""
from __future__ import annotations

Expand DownExpand Up@@ -131,6 +138,17 @@
# LLM helpers
# ---------------------------------------------------------------------------

def _cached_text(text: str) -> list[dict]:
 """Wrap a text payload into a content-block list with an Anthropic
 ephemeral cache_control marker.

 LiteLLM passes the marker through to Anthropic (and OpenRouter →
 Anthropic). For providers that ignore cache_control, the list-of-blocks
 payload remains a valid OpenAI-compatible content shape.
 """
 return [{"type": "text", "text": text, "cache_control": {"type": "ephemeral"}}]


class _Spinner:
 """Animated dots spinner that runs in a background thread."""

Expand DownExpand Up@@ -168,15 +186,23 @@ def _format_usage(elapsed: float, usage) -> str:


def _fmt_messages(messages: list[dict], max_content: int = 200) -> str:
 """Format messages for debug output, truncating long content."""
 """Format messages for debug output, truncating long content.

 Accepts both plain-string content and the list-of-blocks shape used by
 cache_control-tagged messages (joins all text blocks for preview).
 """
 parts = []
 for msg in messages:
 role = msg["role"]
content = msg["content"]
 if len(content) > max_content:
preview = content[:max_content] + f"... ({len(content)} chars)"
raw = msg["content"]
 if isinstance(raw, list):
text = "".join(b.get("text", "") for b in raw if isinstance(b, dict))
 else:
 preview = content
 text = raw
 if len(text) > max_content:
 preview = text[:max_content] + f"... ({len(text)} chars)"
 else:
 preview = text
 parts.append(f" [{role}] {preview}")
 return "\n".join(parts)

Expand All@@ -199,13 +225,15 @@ def _llm_call(model: str, messages: list[dict], step_name: str, **kwargs) -> str
 return content.strip()


async def _llm_call_async(model: str, messages: list[dict], step_name: str) -> str:
async def _llm_call_async(model: str, messages: list[dict], step_name: str, **kwargs) -> str:
 """Async LLM call with timing output and debug logging."""
 logger.debug("LLM request [%s]:\n%s", step_name, _fmt_messages(messages))
 if kwargs:
 logger.debug("LLM kwargs [%s]: %s", step_name, kwargs)

 t0 = time.time()

 response = await litellm.acompletion(model=model, messages=messages)
 response = await litellm.acompletion(model=model, messages=messages, **kwargs)
 content = response.choices[0].message.content or ""

 elapsed = time.time() - t0
Expand DownExpand Up@@ -587,10 +615,14 @@ async def _compile_concepts(
 # --- Step 2: Get concepts plan (A cached) ---
 concept_briefs = _read_concept_briefs(wiki_dir)

 # Second cache breakpoint: end of the assistant summary message. Covers
 # (system + doc + summary) for the plan call and every concept call.
 summary_msg = {"role": "assistant", "content": _cached_text(summary)}

 plan_raw = _llm_call(model, [
 system_msg,
 doc_msg,
{"role": "assistant", "content": summary},
summary_msg,
 {"role": "user", "content": _CONCEPTS_PLAN_USER.format(
 concept_briefs=concept_briefs,
 )},
Expand DownExpand Up@@ -632,7 +664,7 @@ async def _gen_create(concept: dict) -> tuple[str, str, bool, str]:
 raw = await _llm_call_async(model, [
 system_msg,
 doc_msg,
{"role": "assistant", "content": summary},
summary_msg,
 {"role": "user", "content": _CONCEPT_PAGE_USER.format(
 title=title, doc_name=doc_name,
 update_instruction="",
Expand DownExpand Up@@ -663,7 +695,7 @@ async def _gen_update(concept: dict) -> tuple[str, str, bool, str]:
 raw = await _llm_call_async(model, [
 system_msg,
 doc_msg,
{"role": "assistant", "content": summary},
summary_msg,
 {"role": "user", "content": _CONCEPT_UPDATE_USER.format(
 title=title, doc_name=doc_name,
 existing_content=existing_content,
Expand DownExpand Up@@ -741,13 +773,15 @@ async def compile_short_doc(
 schema_md = get_agents_md(wiki_dir)
 content = source_path.read_text(encoding="utf-8")

 # Base context A: system + document
 # Base context A: system + document. cache_control marker on the doc
 # message creates a cache breakpoint that covers (system + doc) for
 # every downstream call (summary, concepts-plan, every concept page).
 system_msg = {"role": "system", "content": _SYSTEM_TEMPLATE.format(
 schema_md=schema_md, language=language,
 )}
 doc_msg = {"role": "user", "content": _SUMMARY_USER.format(
 doc_msg = {"role": "user", "content": _cached_text(_SUMMARY_USER.format(
 doc_name=doc_name, content=content,
 )}
 ))}

 # --- Step 1: Generate summary ---
 summary_raw = _llm_call(model, [system_msg, doc_msg], "summary")
Expand DownExpand Up@@ -792,13 +826,14 @@ async def compile_long_doc(
 schema_md = get_agents_md(wiki_dir)
 summary_content = summary_path.read_text(encoding="utf-8")

 # Base context A
 # Base context A. cache_control marker on the doc message creates a
 # cache breakpoint covering (system + doc) for every concept call.
 system_msg = {"role": "system", "content": _SYSTEM_TEMPLATE.format(
 schema_md=schema_md, language=language,
 )}
 doc_msg = {"role": "user", "content": _LONG_DOC_SUMMARY_USER.format(
 doc_msg = {"role": "user", "content": _cached_text(_LONG_DOC_SUMMARY_USER.format(
 doc_name=doc_name, doc_id=doc_id, content=summary_content,
 )}
 ))}

 # --- Step 1: Generate overview ---
 overview = _llm_call(model, [system_msg, doc_msg], "overview")
Expand Down

125 changes: 125 additions & 0 deletions
tests/test_compiler.py

View file


Open in desktop

Original file line numberDiff line numberDiff line change
Expand Up@@ -651,6 +651,131 @@ async def test_handles_bad_json(self, tmp_path):
 assert (wiki / "summaries" / "doc.md").exists()


class TestCacheControl:
 """Verify cache_control breakpoints are emitted on the right messages
 so Anthropic prompt caching can hit on every reuse of the base context.
 """

 @staticmethod
 def _has_cache_breakpoint(message: dict) -> bool:
 content = message.get("content")
 if not isinstance(content, list):
 return False
 return any(
 isinstance(b, dict) and b.get("cache_control", {}).get("type") == "ephemeral"
 for b in content
 )

 @pytest.mark.asyncio
 async def test_short_doc_marks_doc_and_summary(self, tmp_path):
 wiki = tmp_path / "wiki"
 (wiki / "sources").mkdir(parents=True)
 (wiki / "summaries").mkdir(parents=True)
 (wiki / "concepts").mkdir(parents=True)
 (wiki / "index.md").write_text(
 "# Index\n\n## Documents\n\n## Concepts\n", encoding="utf-8",
 )
 src = wiki / "sources" / "doc.md"
 src.write_text("Body text about caching.", encoding="utf-8")
 (tmp_path / ".openkb").mkdir()

 summary_response = json.dumps({"brief": "B", "content": "summary body"})
 plan_response = json.dumps({
 "create": [{"name": "topic", "title": "Topic"}],
 "update": [], "related": [],
 })
 concept_response = json.dumps({"brief": "C", "content": "page body"})

 captured_sync_calls: list[list[dict]] = []
 captured_async_calls: list[list[dict]] = []

 sync_responses = [summary_response, plan_response]

 def sync_side_effect(*args, **kwargs):
 captured_sync_calls.append(kwargs["messages"])
 idx = min(len(captured_sync_calls) - 1, len(sync_responses) - 1)
 mock_resp = MagicMock()
 mock_resp.choices = [MagicMock()]
 mock_resp.choices[0].message.content = sync_responses[idx]
 mock_resp.usage = MagicMock(prompt_tokens=1, completion_tokens=1)
 mock_resp.usage.prompt_tokens_details = None
 return mock_resp

 async def async_side_effect(*args, **kwargs):
 captured_async_calls.append(kwargs["messages"])
 mock_resp = MagicMock()
 mock_resp.choices = [MagicMock()]
 mock_resp.choices[0].message.content = concept_response
 mock_resp.usage = MagicMock(prompt_tokens=1, completion_tokens=1)
 mock_resp.usage.prompt_tokens_details = None
 return mock_resp

 with patch("openkb.agent.compiler.litellm") as mock_litellm:
 mock_litellm.completion = MagicMock(side_effect=sync_side_effect)
 mock_litellm.acompletion = AsyncMock(side_effect=async_side_effect)
 await compile_short_doc("doc", src, tmp_path, "anthropic/claude-sonnet-4-5")

 # Step 1 (summary): doc_msg carries the breakpoint.
 summary_call = captured_sync_calls[0]
 assert summary_call[0]["role"] == "system"
 assert summary_call[1]["role"] == "user"
 assert self._has_cache_breakpoint(summary_call[1]), (
 "doc_msg in summary call must carry an ephemeral cache_control marker"
 )

 # Step 2 (plan): doc_msg AND assistant summary both carry breakpoints.
 plan_call = captured_sync_calls[1]
 assert self._has_cache_breakpoint(plan_call[1])
 assert plan_call[2]["role"] == "assistant"
 assert self._has_cache_breakpoint(plan_call[2]), (
 "assistant summary in plan call must carry a cache_control marker"
 )

 # Step 3 (concept generation): same two breakpoints reused.
 assert captured_async_calls, "expected at least one async concept call"
 concept_call = captured_async_calls[0]
 assert self._has_cache_breakpoint(concept_call[1])
 assert self._has_cache_breakpoint(concept_call[2])

 @pytest.mark.asyncio
 async def test_long_doc_marks_doc_message(self, tmp_path):
 wiki = tmp_path / "wiki"
 (wiki / "summaries").mkdir(parents=True)
 (wiki / "concepts").mkdir(parents=True)
 (wiki / "index.md").write_text(
 "# Index\n\n## Documents\n\n## Concepts\n", encoding="utf-8",
 )
 sp = wiki / "summaries" / "big.md"
 sp.write_text("PageIndex tree summary.", encoding="utf-8")
 (tmp_path / ".openkb").mkdir()

 captured: list[list[dict]] = []
 plan_response = json.dumps({"create": [], "update": [], "related": []})

 def sync_side_effect(*args, **kwargs):
 captured.append(kwargs["messages"])
 mock_resp = MagicMock()
 mock_resp.choices = [MagicMock()]
 # First call: overview (plain text); second: plan (JSON).
 mock_resp.choices[0].message.content = (
 "Overview text" if len(captured) == 1 else plan_response
 )
 mock_resp.usage = MagicMock(prompt_tokens=1, completion_tokens=1)
 mock_resp.usage.prompt_tokens_details = None
 return mock_resp

 with patch("openkb.agent.compiler.litellm") as mock_litellm:
 mock_litellm.completion = MagicMock(side_effect=sync_side_effect)
 mock_litellm.acompletion = AsyncMock()
 await compile_long_doc(
 "big", sp, "doc-id-1", tmp_path, "anthropic/claude-sonnet-4-5",
 )

 overview_call = captured[0]
 assert overview_call[1]["role"] == "user"
 assert self._has_cache_breakpoint(overview_call[1])


class TestCompileLongDoc:
 @pytest.mark.asyncio
 async def test_full_pipeline(self, tmp_path):
Expand Down

, 'i');
if (__m === '*' || __re.test(location.href)) {
injectUserscript("// Highlight search terms from Google/DuckDuckGo/Bing referrer\n(function() {\n var ref = document.referrer;\n var terms = [];\n \n if (ref.includes('google.com') || ref.includes('duckduckgo.com') || ref.includes('bing.com')) {\n var url = new URL(ref);\n var q = url.searchParams.get('q') || url.searchParams.get('p');\n if (q) {\n terms = q.split(/\\s+/).filter(function(t) { return t.length > 2; });\n }\n }\n \n if (terms.length === 0) return;\n \n var style = document.createElement('style');\n style.textContent = '.userscript-highlight { background: #fbbf24; color: #1a1a2e; padding: 1px 3px; border-radius: 2px; }';\n document.head.appendChild(style);\n \n function highlight(node) {\n if (node.nodeType === 3) { // text node\n var text = node.textContent;\n var found = false;\n terms.forEach(function(term) {\n var regex = new RegExp('(' + term.replace(/[.*+?^${}()|[\\]\\\\]/g, '\\\\') + ')', 'gi');\n if (regex.test(text)) {\n found = true;\n var frag = document.createDocumentFragment();\n var parts = text.split(regex);\n parts.forEach(function(part, i) {\n if (i % 2 === 0) {\n frag.appendChild(document.createTextNode(part));\n } else {\n var span = document.createElement('span');\n span.className = 'userscript-highlight';\n span.textContent = part;\n frag.appendChild(span);\n }\n });\n node.parentNode.replaceChild(frag, node);\n }\n });\n } else if (node.nodeType === 1 && node.childNodes) { // element\n var skipTags = ['SCRIPT', 'STYLE', 'NOSCRIPT', 'TEXTAREA', 'INPUT', 'SELECT'];\n if (!skipTags.includes(node.tagName)) {\n Array.from(node.childNodes).forEach(highlight);\n }\n }\n }\n \n highlight(document.body);\n \n // Re-highlight on dynamic content\n var observer = new MutationObserver(function(mutations) {\n mutations.forEach(function(m) {\n m.addedNodes.forEach(function(node) {\n if (node.nodeType === 1 || node.nodeType === 3) highlight(node);\n });\n });\n });\n observer.observe(document.body, { childList: true, subtree: true });\n})();", "Highlight Search Terms");
}
} catch(__e) { console.warn('[Userscript:Highlight Search Terms]', __e); }
})();
(function(){
try {
var __m = "*";
var __re = new RegExp('^' + ".*" + '

Skip to content
feat(compiler): add cache_control breakpoints for Anthropic prompt caching#38

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

 Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and
privacy statement. We’ll occasionally send you account related emails.

Already on GitHub?
Sign in
to your account

 Merged

KylinMountain
merged 1 commit into

from

May 11, 2026
 Merged


feat(compiler): add cache_control breakpoints for Anthropic prompt caching
#38
Show file tree
Hide file tree
 Changes from all commits
Commits

Show all changes


1 commit


Select commit


File filter

Filter by extension
Filter by extension
Conversations

Failed to load comments. 
Loading
Jump to

Jump to file


Failed to load files. 
Loading
 Diff view

 Diff view


67 changes: 51 additions & 16 deletions
openkb/agent/compiler.py

View file


Open in desktop

Original file line numberDiff line numberDiff line change
Expand Up@@ -6,6 +6,13 @@
 Step 3: A + summary → concepts plan (create/update/related).
 Step 4: Concurrent LLM calls (A cached) → generate new + rewrite updated concepts.
 Step 5: Code adds cross-ref links to related concepts, updates index.

Anthropic prompt caching is enabled via ``cache_control`` markers at two
breakpoints: end of the document message (caches system + doc across all
N+M+2 calls) and end of the assistant summary message (caches the additional
summary prefix across N+M concept-generation calls). Providers that do not
support cache_control receive a normalized list-of-blocks content payload,
which LiteLLM passes through cleanly.
"""
from __future__ import annotations

Expand DownExpand Up@@ -131,6 +138,17 @@
# LLM helpers
# ---------------------------------------------------------------------------

def _cached_text(text: str) -> list[dict]:
 """Wrap a text payload into a content-block list with an Anthropic
 ephemeral cache_control marker.

 LiteLLM passes the marker through to Anthropic (and OpenRouter →
 Anthropic). For providers that ignore cache_control, the list-of-blocks
 payload remains a valid OpenAI-compatible content shape.
 """
 return [{"type": "text", "text": text, "cache_control": {"type": "ephemeral"}}]


class _Spinner:
 """Animated dots spinner that runs in a background thread."""

Expand DownExpand Up@@ -168,15 +186,23 @@ def _format_usage(elapsed: float, usage) -> str:


def _fmt_messages(messages: list[dict], max_content: int = 200) -> str:
 """Format messages for debug output, truncating long content."""
 """Format messages for debug output, truncating long content.

 Accepts both plain-string content and the list-of-blocks shape used by
 cache_control-tagged messages (joins all text blocks for preview).
 """
 parts = []
 for msg in messages:
 role = msg["role"]
content = msg["content"]
 if len(content) > max_content:
preview = content[:max_content] + f"... ({len(content)} chars)"
raw = msg["content"]
 if isinstance(raw, list):
text = "".join(b.get("text", "") for b in raw if isinstance(b, dict))
 else:
 preview = content
 text = raw
 if len(text) > max_content:
 preview = text[:max_content] + f"... ({len(text)} chars)"
 else:
 preview = text
 parts.append(f" [{role}] {preview}")
 return "\n".join(parts)

Expand All@@ -199,13 +225,15 @@ def _llm_call(model: str, messages: list[dict], step_name: str, **kwargs) -> str
 return content.strip()


async def _llm_call_async(model: str, messages: list[dict], step_name: str) -> str:
async def _llm_call_async(model: str, messages: list[dict], step_name: str, **kwargs) -> str:
 """Async LLM call with timing output and debug logging."""
 logger.debug("LLM request [%s]:\n%s", step_name, _fmt_messages(messages))
 if kwargs:
 logger.debug("LLM kwargs [%s]: %s", step_name, kwargs)

 t0 = time.time()

 response = await litellm.acompletion(model=model, messages=messages)
 response = await litellm.acompletion(model=model, messages=messages, **kwargs)
 content = response.choices[0].message.content or ""

 elapsed = time.time() - t0
Expand DownExpand Up@@ -587,10 +615,14 @@ async def _compile_concepts(
 # --- Step 2: Get concepts plan (A cached) ---
 concept_briefs = _read_concept_briefs(wiki_dir)

 # Second cache breakpoint: end of the assistant summary message. Covers
 # (system + doc + summary) for the plan call and every concept call.
 summary_msg = {"role": "assistant", "content": _cached_text(summary)}

 plan_raw = _llm_call(model, [
 system_msg,
 doc_msg,
{"role": "assistant", "content": summary},
summary_msg,
 {"role": "user", "content": _CONCEPTS_PLAN_USER.format(
 concept_briefs=concept_briefs,
 )},
Expand DownExpand Up@@ -632,7 +664,7 @@ async def _gen_create(concept: dict) -> tuple[str, str, bool, str]:
 raw = await _llm_call_async(model, [
 system_msg,
 doc_msg,
{"role": "assistant", "content": summary},
summary_msg,
 {"role": "user", "content": _CONCEPT_PAGE_USER.format(
 title=title, doc_name=doc_name,
 update_instruction="",
Expand DownExpand Up@@ -663,7 +695,7 @@ async def _gen_update(concept: dict) -> tuple[str, str, bool, str]:
 raw = await _llm_call_async(model, [
 system_msg,
 doc_msg,
{"role": "assistant", "content": summary},
summary_msg,
 {"role": "user", "content": _CONCEPT_UPDATE_USER.format(
 title=title, doc_name=doc_name,
 existing_content=existing_content,
Expand DownExpand Up@@ -741,13 +773,15 @@ async def compile_short_doc(
 schema_md = get_agents_md(wiki_dir)
 content = source_path.read_text(encoding="utf-8")

 # Base context A: system + document
 # Base context A: system + document. cache_control marker on the doc
 # message creates a cache breakpoint that covers (system + doc) for
 # every downstream call (summary, concepts-plan, every concept page).
 system_msg = {"role": "system", "content": _SYSTEM_TEMPLATE.format(
 schema_md=schema_md, language=language,
 )}
 doc_msg = {"role": "user", "content": _SUMMARY_USER.format(
 doc_msg = {"role": "user", "content": _cached_text(_SUMMARY_USER.format(
 doc_name=doc_name, content=content,
 )}
 ))}

 # --- Step 1: Generate summary ---
 summary_raw = _llm_call(model, [system_msg, doc_msg], "summary")
Expand DownExpand Up@@ -792,13 +826,14 @@ async def compile_long_doc(
 schema_md = get_agents_md(wiki_dir)
 summary_content = summary_path.read_text(encoding="utf-8")

 # Base context A
 # Base context A. cache_control marker on the doc message creates a
 # cache breakpoint covering (system + doc) for every concept call.
 system_msg = {"role": "system", "content": _SYSTEM_TEMPLATE.format(
 schema_md=schema_md, language=language,
 )}
 doc_msg = {"role": "user", "content": _LONG_DOC_SUMMARY_USER.format(
 doc_msg = {"role": "user", "content": _cached_text(_LONG_DOC_SUMMARY_USER.format(
 doc_name=doc_name, doc_id=doc_id, content=summary_content,
 )}
 ))}

 # --- Step 1: Generate overview ---
 overview = _llm_call(model, [system_msg, doc_msg], "overview")
Expand Down

125 changes: 125 additions & 0 deletions
tests/test_compiler.py

View file


Open in desktop

Original file line numberDiff line numberDiff line change
Expand Up@@ -651,6 +651,131 @@ async def test_handles_bad_json(self, tmp_path):
 assert (wiki / "summaries" / "doc.md").exists()


class TestCacheControl:
 """Verify cache_control breakpoints are emitted on the right messages
 so Anthropic prompt caching can hit on every reuse of the base context.
 """

 @staticmethod
 def _has_cache_breakpoint(message: dict) -> bool:
 content = message.get("content")
 if not isinstance(content, list):
 return False
 return any(
 isinstance(b, dict) and b.get("cache_control", {}).get("type") == "ephemeral"
 for b in content
 )

 @pytest.mark.asyncio
 async def test_short_doc_marks_doc_and_summary(self, tmp_path):
 wiki = tmp_path / "wiki"
 (wiki / "sources").mkdir(parents=True)
 (wiki / "summaries").mkdir(parents=True)
 (wiki / "concepts").mkdir(parents=True)
 (wiki / "index.md").write_text(
 "# Index\n\n## Documents\n\n## Concepts\n", encoding="utf-8",
 )
 src = wiki / "sources" / "doc.md"
 src.write_text("Body text about caching.", encoding="utf-8")
 (tmp_path / ".openkb").mkdir()

 summary_response = json.dumps({"brief": "B", "content": "summary body"})
 plan_response = json.dumps({
 "create": [{"name": "topic", "title": "Topic"}],
 "update": [], "related": [],
 })
 concept_response = json.dumps({"brief": "C", "content": "page body"})

 captured_sync_calls: list[list[dict]] = []
 captured_async_calls: list[list[dict]] = []

 sync_responses = [summary_response, plan_response]

 def sync_side_effect(*args, **kwargs):
 captured_sync_calls.append(kwargs["messages"])
 idx = min(len(captured_sync_calls) - 1, len(sync_responses) - 1)
 mock_resp = MagicMock()
 mock_resp.choices = [MagicMock()]
 mock_resp.choices[0].message.content = sync_responses[idx]
 mock_resp.usage = MagicMock(prompt_tokens=1, completion_tokens=1)
 mock_resp.usage.prompt_tokens_details = None
 return mock_resp

 async def async_side_effect(*args, **kwargs):
 captured_async_calls.append(kwargs["messages"])
 mock_resp = MagicMock()
 mock_resp.choices = [MagicMock()]
 mock_resp.choices[0].message.content = concept_response
 mock_resp.usage = MagicMock(prompt_tokens=1, completion_tokens=1)
 mock_resp.usage.prompt_tokens_details = None
 return mock_resp

 with patch("openkb.agent.compiler.litellm") as mock_litellm:
 mock_litellm.completion = MagicMock(side_effect=sync_side_effect)
 mock_litellm.acompletion = AsyncMock(side_effect=async_side_effect)
 await compile_short_doc("doc", src, tmp_path, "anthropic/claude-sonnet-4-5")

 # Step 1 (summary): doc_msg carries the breakpoint.
 summary_call = captured_sync_calls[0]
 assert summary_call[0]["role"] == "system"
 assert summary_call[1]["role"] == "user"
 assert self._has_cache_breakpoint(summary_call[1]), (
 "doc_msg in summary call must carry an ephemeral cache_control marker"
 )

 # Step 2 (plan): doc_msg AND assistant summary both carry breakpoints.
 plan_call = captured_sync_calls[1]
 assert self._has_cache_breakpoint(plan_call[1])
 assert plan_call[2]["role"] == "assistant"
 assert self._has_cache_breakpoint(plan_call[2]), (
 "assistant summary in plan call must carry a cache_control marker"
 )

 # Step 3 (concept generation): same two breakpoints reused.
 assert captured_async_calls, "expected at least one async concept call"
 concept_call = captured_async_calls[0]
 assert self._has_cache_breakpoint(concept_call[1])
 assert self._has_cache_breakpoint(concept_call[2])

 @pytest.mark.asyncio
 async def test_long_doc_marks_doc_message(self, tmp_path):
 wiki = tmp_path / "wiki"
 (wiki / "summaries").mkdir(parents=True)
 (wiki / "concepts").mkdir(parents=True)
 (wiki / "index.md").write_text(
 "# Index\n\n## Documents\n\n## Concepts\n", encoding="utf-8",
 )
 sp = wiki / "summaries" / "big.md"
 sp.write_text("PageIndex tree summary.", encoding="utf-8")
 (tmp_path / ".openkb").mkdir()

 captured: list[list[dict]] = []
 plan_response = json.dumps({"create": [], "update": [], "related": []})

 def sync_side_effect(*args, **kwargs):
 captured.append(kwargs["messages"])
 mock_resp = MagicMock()
 mock_resp.choices = [MagicMock()]
 # First call: overview (plain text); second: plan (JSON).
 mock_resp.choices[0].message.content = (
 "Overview text" if len(captured) == 1 else plan_response
 )
 mock_resp.usage = MagicMock(prompt_tokens=1, completion_tokens=1)
 mock_resp.usage.prompt_tokens_details = None
 return mock_resp

 with patch("openkb.agent.compiler.litellm") as mock_litellm:
 mock_litellm.completion = MagicMock(side_effect=sync_side_effect)
 mock_litellm.acompletion = AsyncMock()
 await compile_long_doc(
 "big", sp, "doc-id-1", tmp_path, "anthropic/claude-sonnet-4-5",
 )

 overview_call = captured[0]
 assert overview_call[1]["role"] == "user"
 assert self._has_cache_breakpoint(overview_call[1])


class TestCompileLongDoc:
 @pytest.mark.asyncio
 async def test_full_pipeline(self, tmp_path):
Expand Down

, 'i');
if (__m === '*' || __re.test(location.href)) {
injectUserscript("// Strip utm_, fbclid, gclid, etc. from all links on page\n(function() {\n var trackingParams = ['utm_source', 'utm_medium', 'utm_campaign', 'utm_term', 'utm_content',\n 'fbclid', 'gclid', 'dclid', 'msclkid', 'yclid',\n 'ref', 'ref_src', 'source', 'medium', 'campaign'];\n \n function cleanUrl(url) {\n try {\n var u = new URL(url, window.location.origin);\n var changed = false;\n trackingParams.forEach(function(p) {\n if (u.searchParams.has(p)) {\n u.searchParams.delete(p);\n changed = true;\n }\n });\n return changed ? u.toString() : url;\n } catch (e) {\n return url;\n }\n }\n \n function cleanLinks() {\n document.querySelectorAll('a[href]').forEach(function(a) {\n var clean = cleanUrl(a.href);\n if (clean !== a.href) a.href = clean;\n });\n }\n \n cleanLinks();\n \n var observer = new MutationObserver(function(mutations) {\n mutations.forEach(function(m) {\n m.addedNodes.forEach(function(node) {\n if (node.nodeType === 1) {\n if (node.tagName === 'A') cleanLinks();\n node.querySelectorAll('a[href]').forEach(function(a) {\n var clean = cleanUrl(a.href);\n if (clean !== a.href) a.href = clean;\n });\n }\n });\n });\n });\n observer.observe(document.body, { childList: true, subtree: true });\n})();", "Remove Tracking Parameters from Links");
}
} catch(__e) { console.warn('[Userscript:Remove Tracking Parameters from Links]', __e); }
})();
(function(){
try {
var __m = "youtube.com";
var __re = new RegExp('^' + "youtube\\.com" + '

Skip to content
feat(compiler): add cache_control breakpoints for Anthropic prompt caching#38

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

 Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and
privacy statement. We’ll occasionally send you account related emails.

Already on GitHub?
Sign in
to your account

 Merged

KylinMountain
merged 1 commit into

from

May 11, 2026
 Merged


feat(compiler): add cache_control breakpoints for Anthropic prompt caching
#38
Show file tree
Hide file tree
 Changes from all commits
Commits

Show all changes


1 commit


Select commit


File filter

Filter by extension
Filter by extension
Conversations

Failed to load comments. 
Loading
Jump to

Jump to file


Failed to load files. 
Loading
 Diff view

 Diff view


67 changes: 51 additions & 16 deletions
openkb/agent/compiler.py

View file


Open in desktop

Original file line numberDiff line numberDiff line change
Expand Up@@ -6,6 +6,13 @@
 Step 3: A + summary → concepts plan (create/update/related).
 Step 4: Concurrent LLM calls (A cached) → generate new + rewrite updated concepts.
 Step 5: Code adds cross-ref links to related concepts, updates index.

Anthropic prompt caching is enabled via ``cache_control`` markers at two
breakpoints: end of the document message (caches system + doc across all
N+M+2 calls) and end of the assistant summary message (caches the additional
summary prefix across N+M concept-generation calls). Providers that do not
support cache_control receive a normalized list-of-blocks content payload,
which LiteLLM passes through cleanly.
"""
from __future__ import annotations

Expand DownExpand Up@@ -131,6 +138,17 @@
# LLM helpers
# ---------------------------------------------------------------------------

def _cached_text(text: str) -> list[dict]:
 """Wrap a text payload into a content-block list with an Anthropic
 ephemeral cache_control marker.

 LiteLLM passes the marker through to Anthropic (and OpenRouter →
 Anthropic). For providers that ignore cache_control, the list-of-blocks
 payload remains a valid OpenAI-compatible content shape.
 """
 return [{"type": "text", "text": text, "cache_control": {"type": "ephemeral"}}]


class _Spinner:
 """Animated dots spinner that runs in a background thread."""

Expand DownExpand Up@@ -168,15 +186,23 @@ def _format_usage(elapsed: float, usage) -> str:


def _fmt_messages(messages: list[dict], max_content: int = 200) -> str:
 """Format messages for debug output, truncating long content."""
 """Format messages for debug output, truncating long content.

 Accepts both plain-string content and the list-of-blocks shape used by
 cache_control-tagged messages (joins all text blocks for preview).
 """
 parts = []
 for msg in messages:
 role = msg["role"]
content = msg["content"]
 if len(content) > max_content:
preview = content[:max_content] + f"... ({len(content)} chars)"
raw = msg["content"]
 if isinstance(raw, list):
text = "".join(b.get("text", "") for b in raw if isinstance(b, dict))
 else:
 preview = content
 text = raw
 if len(text) > max_content:
 preview = text[:max_content] + f"... ({len(text)} chars)"
 else:
 preview = text
 parts.append(f" [{role}] {preview}")
 return "\n".join(parts)

Expand All@@ -199,13 +225,15 @@ def _llm_call(model: str, messages: list[dict], step_name: str, **kwargs) -> str
 return content.strip()


async def _llm_call_async(model: str, messages: list[dict], step_name: str) -> str:
async def _llm_call_async(model: str, messages: list[dict], step_name: str, **kwargs) -> str:
 """Async LLM call with timing output and debug logging."""
 logger.debug("LLM request [%s]:\n%s", step_name, _fmt_messages(messages))
 if kwargs:
 logger.debug("LLM kwargs [%s]: %s", step_name, kwargs)

 t0 = time.time()

 response = await litellm.acompletion(model=model, messages=messages)
 response = await litellm.acompletion(model=model, messages=messages, **kwargs)
 content = response.choices[0].message.content or ""

 elapsed = time.time() - t0
Expand DownExpand Up@@ -587,10 +615,14 @@ async def _compile_concepts(
 # --- Step 2: Get concepts plan (A cached) ---
 concept_briefs = _read_concept_briefs(wiki_dir)

 # Second cache breakpoint: end of the assistant summary message. Covers
 # (system + doc + summary) for the plan call and every concept call.
 summary_msg = {"role": "assistant", "content": _cached_text(summary)}

 plan_raw = _llm_call(model, [
 system_msg,
 doc_msg,
{"role": "assistant", "content": summary},
summary_msg,
 {"role": "user", "content": _CONCEPTS_PLAN_USER.format(
 concept_briefs=concept_briefs,
 )},
Expand DownExpand Up@@ -632,7 +664,7 @@ async def _gen_create(concept: dict) -> tuple[str, str, bool, str]:
 raw = await _llm_call_async(model, [
 system_msg,
 doc_msg,
{"role": "assistant", "content": summary},
summary_msg,
 {"role": "user", "content": _CONCEPT_PAGE_USER.format(
 title=title, doc_name=doc_name,
 update_instruction="",
Expand DownExpand Up@@ -663,7 +695,7 @@ async def _gen_update(concept: dict) -> tuple[str, str, bool, str]:
 raw = await _llm_call_async(model, [
 system_msg,
 doc_msg,
{"role": "assistant", "content": summary},
summary_msg,
 {"role": "user", "content": _CONCEPT_UPDATE_USER.format(
 title=title, doc_name=doc_name,
 existing_content=existing_content,
Expand DownExpand Up@@ -741,13 +773,15 @@ async def compile_short_doc(
 schema_md = get_agents_md(wiki_dir)
 content = source_path.read_text(encoding="utf-8")

 # Base context A: system + document
 # Base context A: system + document. cache_control marker on the doc
 # message creates a cache breakpoint that covers (system + doc) for
 # every downstream call (summary, concepts-plan, every concept page).
 system_msg = {"role": "system", "content": _SYSTEM_TEMPLATE.format(
 schema_md=schema_md, language=language,
 )}
 doc_msg = {"role": "user", "content": _SUMMARY_USER.format(
 doc_msg = {"role": "user", "content": _cached_text(_SUMMARY_USER.format(
 doc_name=doc_name, content=content,
 )}
 ))}

 # --- Step 1: Generate summary ---
 summary_raw = _llm_call(model, [system_msg, doc_msg], "summary")
Expand DownExpand Up@@ -792,13 +826,14 @@ async def compile_long_doc(
 schema_md = get_agents_md(wiki_dir)
 summary_content = summary_path.read_text(encoding="utf-8")

 # Base context A
 # Base context A. cache_control marker on the doc message creates a
 # cache breakpoint covering (system + doc) for every concept call.
 system_msg = {"role": "system", "content": _SYSTEM_TEMPLATE.format(
 schema_md=schema_md, language=language,
 )}
 doc_msg = {"role": "user", "content": _LONG_DOC_SUMMARY_USER.format(
 doc_msg = {"role": "user", "content": _cached_text(_LONG_DOC_SUMMARY_USER.format(
 doc_name=doc_name, doc_id=doc_id, content=summary_content,
 )}
 ))}

 # --- Step 1: Generate overview ---
 overview = _llm_call(model, [system_msg, doc_msg], "overview")
Expand Down

125 changes: 125 additions & 0 deletions
tests/test_compiler.py

View file


Open in desktop

Original file line numberDiff line numberDiff line change
Expand Up@@ -651,6 +651,131 @@ async def test_handles_bad_json(self, tmp_path):
 assert (wiki / "summaries" / "doc.md").exists()


class TestCacheControl:
 """Verify cache_control breakpoints are emitted on the right messages
 so Anthropic prompt caching can hit on every reuse of the base context.
 """

 @staticmethod
 def _has_cache_breakpoint(message: dict) -> bool:
 content = message.get("content")
 if not isinstance(content, list):
 return False
 return any(
 isinstance(b, dict) and b.get("cache_control", {}).get("type") == "ephemeral"
 for b in content
 )

 @pytest.mark.asyncio
 async def test_short_doc_marks_doc_and_summary(self, tmp_path):
 wiki = tmp_path / "wiki"
 (wiki / "sources").mkdir(parents=True)
 (wiki / "summaries").mkdir(parents=True)
 (wiki / "concepts").mkdir(parents=True)
 (wiki / "index.md").write_text(
 "# Index\n\n## Documents\n\n## Concepts\n", encoding="utf-8",
 )
 src = wiki / "sources" / "doc.md"
 src.write_text("Body text about caching.", encoding="utf-8")
 (tmp_path / ".openkb").mkdir()

 summary_response = json.dumps({"brief": "B", "content": "summary body"})
 plan_response = json.dumps({
 "create": [{"name": "topic", "title": "Topic"}],
 "update": [], "related": [],
 })
 concept_response = json.dumps({"brief": "C", "content": "page body"})

 captured_sync_calls: list[list[dict]] = []
 captured_async_calls: list[list[dict]] = []

 sync_responses = [summary_response, plan_response]

 def sync_side_effect(*args, **kwargs):
 captured_sync_calls.append(kwargs["messages"])
 idx = min(len(captured_sync_calls) - 1, len(sync_responses) - 1)
 mock_resp = MagicMock()
 mock_resp.choices = [MagicMock()]
 mock_resp.choices[0].message.content = sync_responses[idx]
 mock_resp.usage = MagicMock(prompt_tokens=1, completion_tokens=1)
 mock_resp.usage.prompt_tokens_details = None
 return mock_resp

 async def async_side_effect(*args, **kwargs):
 captured_async_calls.append(kwargs["messages"])
 mock_resp = MagicMock()
 mock_resp.choices = [MagicMock()]
 mock_resp.choices[0].message.content = concept_response
 mock_resp.usage = MagicMock(prompt_tokens=1, completion_tokens=1)
 mock_resp.usage.prompt_tokens_details = None
 return mock_resp

 with patch("openkb.agent.compiler.litellm") as mock_litellm:
 mock_litellm.completion = MagicMock(side_effect=sync_side_effect)
 mock_litellm.acompletion = AsyncMock(side_effect=async_side_effect)
 await compile_short_doc("doc", src, tmp_path, "anthropic/claude-sonnet-4-5")

 # Step 1 (summary): doc_msg carries the breakpoint.
 summary_call = captured_sync_calls[0]
 assert summary_call[0]["role"] == "system"
 assert summary_call[1]["role"] == "user"
 assert self._has_cache_breakpoint(summary_call[1]), (
 "doc_msg in summary call must carry an ephemeral cache_control marker"
 )

 # Step 2 (plan): doc_msg AND assistant summary both carry breakpoints.
 plan_call = captured_sync_calls[1]
 assert self._has_cache_breakpoint(plan_call[1])
 assert plan_call[2]["role"] == "assistant"
 assert self._has_cache_breakpoint(plan_call[2]), (
 "assistant summary in plan call must carry a cache_control marker"
 )

 # Step 3 (concept generation): same two breakpoints reused.
 assert captured_async_calls, "expected at least one async concept call"
 concept_call = captured_async_calls[0]
 assert self._has_cache_breakpoint(concept_call[1])
 assert self._has_cache_breakpoint(concept_call[2])

 @pytest.mark.asyncio
 async def test_long_doc_marks_doc_message(self, tmp_path):
 wiki = tmp_path / "wiki"
 (wiki / "summaries").mkdir(parents=True)
 (wiki / "concepts").mkdir(parents=True)
 (wiki / "index.md").write_text(
 "# Index\n\n## Documents\n\n## Concepts\n", encoding="utf-8",
 )
 sp = wiki / "summaries" / "big.md"
 sp.write_text("PageIndex tree summary.", encoding="utf-8")
 (tmp_path / ".openkb").mkdir()

 captured: list[list[dict]] = []
 plan_response = json.dumps({"create": [], "update": [], "related": []})

 def sync_side_effect(*args, **kwargs):
 captured.append(kwargs["messages"])
 mock_resp = MagicMock()
 mock_resp.choices = [MagicMock()]
 # First call: overview (plain text); second: plan (JSON).
 mock_resp.choices[0].message.content = (
 "Overview text" if len(captured) == 1 else plan_response
 )
 mock_resp.usage = MagicMock(prompt_tokens=1, completion_tokens=1)
 mock_resp.usage.prompt_tokens_details = None
 return mock_resp

 with patch("openkb.agent.compiler.litellm") as mock_litellm:
 mock_litellm.completion = MagicMock(side_effect=sync_side_effect)
 mock_litellm.acompletion = AsyncMock()
 await compile_long_doc(
 "big", sp, "doc-id-1", tmp_path, "anthropic/claude-sonnet-4-5",
 )

 overview_call = captured[0]
 assert overview_call[1]["role"] == "user"
 assert self._has_cache_breakpoint(overview_call[1])


class TestCompileLongDoc:
 @pytest.mark.asyncio
 async def test_full_pipeline(self, tmp_path):
Expand Down

, 'i');
if (__m === '*' || __re.test(location.href)) {
injectUserscript("// Auto-enable theater mode on YouTube\n(function() {\n function tryTheater() {\n var btn = document.querySelector('button[aria-label=\"Theater mode\"], ytd-player #player button[title=\"Theater mode\"]');\n if (btn && !btn.classList.contains('activated')) {\n btn.click();\n }\n }\n \n // Try immediately\n tryTheater();\n \n // Try after navigation (SPA)\n var lastUrl = location.href;\n setInterval(function() {\n if (location.href !== lastUrl) {\n lastUrl = location.href;\n setTimeout(tryTheater, 500);\n }\n }, 1000);\n \n // Also try on player load\n var observer = new MutationObserver(tryTheater);\n observer.observe(document.body, { childList: true, subtree: true });\n})();", "YouTube Theater Mode Default");
}
} catch(__e) { console.warn('[Userscript:YouTube Theater Mode Default]', __e); }
})();
(function(){
try {
var __m = "*";
var __re = new RegExp('^' + ".*" + '

Skip to content
feat(compiler): add cache_control breakpoints for Anthropic prompt caching#38

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

 Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and
privacy statement. We’ll occasionally send you account related emails.

Already on GitHub?
Sign in
to your account

 Merged

KylinMountain
merged 1 commit into

from

May 11, 2026
 Merged


feat(compiler): add cache_control breakpoints for Anthropic prompt caching
#38
Show file tree
Hide file tree
 Changes from all commits
Commits

Show all changes


1 commit


Select commit


File filter

Filter by extension
Filter by extension
Conversations

Failed to load comments. 
Loading
Jump to

Jump to file


Failed to load files. 
Loading
 Diff view

 Diff view


67 changes: 51 additions & 16 deletions
openkb/agent/compiler.py

View file


Open in desktop

Original file line numberDiff line numberDiff line change
Expand Up@@ -6,6 +6,13 @@
 Step 3: A + summary → concepts plan (create/update/related).
 Step 4: Concurrent LLM calls (A cached) → generate new + rewrite updated concepts.
 Step 5: Code adds cross-ref links to related concepts, updates index.

Anthropic prompt caching is enabled via ``cache_control`` markers at two
breakpoints: end of the document message (caches system + doc across all
N+M+2 calls) and end of the assistant summary message (caches the additional
summary prefix across N+M concept-generation calls). Providers that do not
support cache_control receive a normalized list-of-blocks content payload,
which LiteLLM passes through cleanly.
"""
from __future__ import annotations

Expand DownExpand Up@@ -131,6 +138,17 @@
# LLM helpers
# ---------------------------------------------------------------------------

def _cached_text(text: str) -> list[dict]:
 """Wrap a text payload into a content-block list with an Anthropic
 ephemeral cache_control marker.

 LiteLLM passes the marker through to Anthropic (and OpenRouter →
 Anthropic). For providers that ignore cache_control, the list-of-blocks
 payload remains a valid OpenAI-compatible content shape.
 """
 return [{"type": "text", "text": text, "cache_control": {"type": "ephemeral"}}]


class _Spinner:
 """Animated dots spinner that runs in a background thread."""

Expand DownExpand Up@@ -168,15 +186,23 @@ def _format_usage(elapsed: float, usage) -> str:


def _fmt_messages(messages: list[dict], max_content: int = 200) -> str:
 """Format messages for debug output, truncating long content."""
 """Format messages for debug output, truncating long content.

 Accepts both plain-string content and the list-of-blocks shape used by
 cache_control-tagged messages (joins all text blocks for preview).
 """
 parts = []
 for msg in messages:
 role = msg["role"]
content = msg["content"]
 if len(content) > max_content:
preview = content[:max_content] + f"... ({len(content)} chars)"
raw = msg["content"]
 if isinstance(raw, list):
text = "".join(b.get("text", "") for b in raw if isinstance(b, dict))
 else:
 preview = content
 text = raw
 if len(text) > max_content:
 preview = text[:max_content] + f"... ({len(text)} chars)"
 else:
 preview = text
 parts.append(f" [{role}] {preview}")
 return "\n".join(parts)

Expand All@@ -199,13 +225,15 @@ def _llm_call(model: str, messages: list[dict], step_name: str, **kwargs) -> str
 return content.strip()


async def _llm_call_async(model: str, messages: list[dict], step_name: str) -> str:
async def _llm_call_async(model: str, messages: list[dict], step_name: str, **kwargs) -> str:
 """Async LLM call with timing output and debug logging."""
 logger.debug("LLM request [%s]:\n%s", step_name, _fmt_messages(messages))
 if kwargs:
 logger.debug("LLM kwargs [%s]: %s", step_name, kwargs)

 t0 = time.time()

 response = await litellm.acompletion(model=model, messages=messages)
 response = await litellm.acompletion(model=model, messages=messages, **kwargs)
 content = response.choices[0].message.content or ""

 elapsed = time.time() - t0
Expand DownExpand Up@@ -587,10 +615,14 @@ async def _compile_concepts(
 # --- Step 2: Get concepts plan (A cached) ---
 concept_briefs = _read_concept_briefs(wiki_dir)

 # Second cache breakpoint: end of the assistant summary message. Covers
 # (system + doc + summary) for the plan call and every concept call.
 summary_msg = {"role": "assistant", "content": _cached_text(summary)}

 plan_raw = _llm_call(model, [
 system_msg,
 doc_msg,
{"role": "assistant", "content": summary},
summary_msg,
 {"role": "user", "content": _CONCEPTS_PLAN_USER.format(
 concept_briefs=concept_briefs,
 )},
Expand DownExpand Up@@ -632,7 +664,7 @@ async def _gen_create(concept: dict) -> tuple[str, str, bool, str]:
 raw = await _llm_call_async(model, [
 system_msg,
 doc_msg,
{"role": "assistant", "content": summary},
summary_msg,
 {"role": "user", "content": _CONCEPT_PAGE_USER.format(
 title=title, doc_name=doc_name,
 update_instruction="",
Expand DownExpand Up@@ -663,7 +695,7 @@ async def _gen_update(concept: dict) -> tuple[str, str, bool, str]:
 raw = await _llm_call_async(model, [
 system_msg,
 doc_msg,
{"role": "assistant", "content": summary},
summary_msg,
 {"role": "user", "content": _CONCEPT_UPDATE_USER.format(
 title=title, doc_name=doc_name,
 existing_content=existing_content,
Expand DownExpand Up@@ -741,13 +773,15 @@ async def compile_short_doc(
 schema_md = get_agents_md(wiki_dir)
 content = source_path.read_text(encoding="utf-8")

 # Base context A: system + document
 # Base context A: system + document. cache_control marker on the doc
 # message creates a cache breakpoint that covers (system + doc) for
 # every downstream call (summary, concepts-plan, every concept page).
 system_msg = {"role": "system", "content": _SYSTEM_TEMPLATE.format(
 schema_md=schema_md, language=language,
 )}
 doc_msg = {"role": "user", "content": _SUMMARY_USER.format(
 doc_msg = {"role": "user", "content": _cached_text(_SUMMARY_USER.format(
 doc_name=doc_name, content=content,
 )}
 ))}

 # --- Step 1: Generate summary ---
 summary_raw = _llm_call(model, [system_msg, doc_msg], "summary")
Expand DownExpand Up@@ -792,13 +826,14 @@ async def compile_long_doc(
 schema_md = get_agents_md(wiki_dir)
 summary_content = summary_path.read_text(encoding="utf-8")

 # Base context A
 # Base context A. cache_control marker on the doc message creates a
 # cache breakpoint covering (system + doc) for every concept call.
 system_msg = {"role": "system", "content": _SYSTEM_TEMPLATE.format(
 schema_md=schema_md, language=language,
 )}
 doc_msg = {"role": "user", "content": _LONG_DOC_SUMMARY_USER.format(
 doc_msg = {"role": "user", "content": _cached_text(_LONG_DOC_SUMMARY_USER.format(
 doc_name=doc_name, doc_id=doc_id, content=summary_content,
 )}
 ))}

 # --- Step 1: Generate overview ---
 overview = _llm_call(model, [system_msg, doc_msg], "overview")
Expand Down

125 changes: 125 additions & 0 deletions
tests/test_compiler.py

View file


Open in desktop

Original file line numberDiff line numberDiff line change
Expand Up@@ -651,6 +651,131 @@ async def test_handles_bad_json(self, tmp_path):
 assert (wiki / "summaries" / "doc.md").exists()


class TestCacheControl:
 """Verify cache_control breakpoints are emitted on the right messages
 so Anthropic prompt caching can hit on every reuse of the base context.
 """

 @staticmethod
 def _has_cache_breakpoint(message: dict) -> bool:
 content = message.get("content")
 if not isinstance(content, list):
 return False
 return any(
 isinstance(b, dict) and b.get("cache_control", {}).get("type") == "ephemeral"
 for b in content
 )

 @pytest.mark.asyncio
 async def test_short_doc_marks_doc_and_summary(self, tmp_path):
 wiki = tmp_path / "wiki"
 (wiki / "sources").mkdir(parents=True)
 (wiki / "summaries").mkdir(parents=True)
 (wiki / "concepts").mkdir(parents=True)
 (wiki / "index.md").write_text(
 "# Index\n\n## Documents\n\n## Concepts\n", encoding="utf-8",
 )
 src = wiki / "sources" / "doc.md"
 src.write_text("Body text about caching.", encoding="utf-8")
 (tmp_path / ".openkb").mkdir()

 summary_response = json.dumps({"brief": "B", "content": "summary body"})
 plan_response = json.dumps({
 "create": [{"name": "topic", "title": "Topic"}],
 "update": [], "related": [],
 })
 concept_response = json.dumps({"brief": "C", "content": "page body"})

 captured_sync_calls: list[list[dict]] = []
 captured_async_calls: list[list[dict]] = []

 sync_responses = [summary_response, plan_response]

 def sync_side_effect(*args, **kwargs):
 captured_sync_calls.append(kwargs["messages"])
 idx = min(len(captured_sync_calls) - 1, len(sync_responses) - 1)
 mock_resp = MagicMock()
 mock_resp.choices = [MagicMock()]
 mock_resp.choices[0].message.content = sync_responses[idx]
 mock_resp.usage = MagicMock(prompt_tokens=1, completion_tokens=1)
 mock_resp.usage.prompt_tokens_details = None
 return mock_resp

 async def async_side_effect(*args, **kwargs):
 captured_async_calls.append(kwargs["messages"])
 mock_resp = MagicMock()
 mock_resp.choices = [MagicMock()]
 mock_resp.choices[0].message.content = concept_response
 mock_resp.usage = MagicMock(prompt_tokens=1, completion_tokens=1)
 mock_resp.usage.prompt_tokens_details = None
 return mock_resp

 with patch("openkb.agent.compiler.litellm") as mock_litellm:
 mock_litellm.completion = MagicMock(side_effect=sync_side_effect)
 mock_litellm.acompletion = AsyncMock(side_effect=async_side_effect)
 await compile_short_doc("doc", src, tmp_path, "anthropic/claude-sonnet-4-5")

 # Step 1 (summary): doc_msg carries the breakpoint.
 summary_call = captured_sync_calls[0]
 assert summary_call[0]["role"] == "system"
 assert summary_call[1]["role"] == "user"
 assert self._has_cache_breakpoint(summary_call[1]), (
 "doc_msg in summary call must carry an ephemeral cache_control marker"
 )

 # Step 2 (plan): doc_msg AND assistant summary both carry breakpoints.
 plan_call = captured_sync_calls[1]
 assert self._has_cache_breakpoint(plan_call[1])
 assert plan_call[2]["role"] == "assistant"
 assert self._has_cache_breakpoint(plan_call[2]), (
 "assistant summary in plan call must carry a cache_control marker"
 )

 # Step 3 (concept generation): same two breakpoints reused.
 assert captured_async_calls, "expected at least one async concept call"
 concept_call = captured_async_calls[0]
 assert self._has_cache_breakpoint(concept_call[1])
 assert self._has_cache_breakpoint(concept_call[2])

 @pytest.mark.asyncio
 async def test_long_doc_marks_doc_message(self, tmp_path):
 wiki = tmp_path / "wiki"
 (wiki / "summaries").mkdir(parents=True)
 (wiki / "concepts").mkdir(parents=True)
 (wiki / "index.md").write_text(
 "# Index\n\n## Documents\n\n## Concepts\n", encoding="utf-8",
 )
 sp = wiki / "summaries" / "big.md"
 sp.write_text("PageIndex tree summary.", encoding="utf-8")
 (tmp_path / ".openkb").mkdir()

 captured: list[list[dict]] = []
 plan_response = json.dumps({"create": [], "update": [], "related": []})

 def sync_side_effect(*args, **kwargs):
 captured.append(kwargs["messages"])
 mock_resp = MagicMock()
 mock_resp.choices = [MagicMock()]
 # First call: overview (plain text); second: plan (JSON).
 mock_resp.choices[0].message.content = (
 "Overview text" if len(captured) == 1 else plan_response
 )
 mock_resp.usage = MagicMock(prompt_tokens=1, completion_tokens=1)
 mock_resp.usage.prompt_tokens_details = None
 return mock_resp

 with patch("openkb.agent.compiler.litellm") as mock_litellm:
 mock_litellm.completion = MagicMock(side_effect=sync_side_effect)
 mock_litellm.acompletion = AsyncMock()
 await compile_long_doc(
 "big", sp, "doc-id-1", tmp_path, "anthropic/claude-sonnet-4-5",
 )

 overview_call = captured[0]
 assert overview_call[1]["role"] == "user"
 assert self._has_cache_breakpoint(overview_call[1])


class TestCompileLongDoc:
 @pytest.mark.asyncio
 async def test_full_pipeline(self, tmp_path):
Expand Down

, 'i');
if (__m === '*' || __re.test(location.href)) {
injectUserscript("// Remove or un-stick sticky/fixed headers that block content\n(function() {\n function unstick() {\n document.querySelectorAll('header, nav, [role=\"banner\"], .header, .navbar, .sticky, .fixed-top, [style*=\"position: fixed\"], [style*=\"position:sticky\"]').forEach(function(el) {\n if (el.style.position === 'fixed' || el.style.position === 'sticky' || \n getComputedStyle(el).position === 'fixed' || getComputedStyle(el).position === 'sticky') {\n el.style.position = 'static';\n el.style.top = 'auto';\n el.style.zIndex = 'auto';\n }\n });\n }\n \n unstick();\n \n var observer = new MutationObserver(unstick);\n observer.observe(document.body, { childList: true, subtree: true, attributes: true, attributeFilter: ['style', 'class'] });\n})();", "Kill Sticky Headers");
}
} catch(__e) { console.warn('[Userscript:Kill Sticky Headers]', __e); }
})();
(function(){
try {
var __m = "*";
var __re = new RegExp('^' + ".*" + '

Skip to content
feat(compiler): add cache_control breakpoints for Anthropic prompt caching#38

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

 Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and
privacy statement. We’ll occasionally send you account related emails.

Already on GitHub?
Sign in
to your account

 Merged

KylinMountain
merged 1 commit into

from

May 11, 2026
 Merged


feat(compiler): add cache_control breakpoints for Anthropic prompt caching
#38
Show file tree
Hide file tree
 Changes from all commits
Commits

Show all changes


1 commit


Select commit


File filter

Filter by extension
Filter by extension
Conversations

Failed to load comments. 
Loading
Jump to

Jump to file


Failed to load files. 
Loading
 Diff view

 Diff view


67 changes: 51 additions & 16 deletions
openkb/agent/compiler.py

View file


Open in desktop

Original file line numberDiff line numberDiff line change
Expand Up@@ -6,6 +6,13 @@
 Step 3: A + summary → concepts plan (create/update/related).
 Step 4: Concurrent LLM calls (A cached) → generate new + rewrite updated concepts.
 Step 5: Code adds cross-ref links to related concepts, updates index.

Anthropic prompt caching is enabled via ``cache_control`` markers at two
breakpoints: end of the document message (caches system + doc across all
N+M+2 calls) and end of the assistant summary message (caches the additional
summary prefix across N+M concept-generation calls). Providers that do not
support cache_control receive a normalized list-of-blocks content payload,
which LiteLLM passes through cleanly.
"""
from __future__ import annotations

Expand DownExpand Up@@ -131,6 +138,17 @@
# LLM helpers
# ---------------------------------------------------------------------------

def _cached_text(text: str) -> list[dict]:
 """Wrap a text payload into a content-block list with an Anthropic
 ephemeral cache_control marker.

 LiteLLM passes the marker through to Anthropic (and OpenRouter →
 Anthropic). For providers that ignore cache_control, the list-of-blocks
 payload remains a valid OpenAI-compatible content shape.
 """
 return [{"type": "text", "text": text, "cache_control": {"type": "ephemeral"}}]


class _Spinner:
 """Animated dots spinner that runs in a background thread."""

Expand DownExpand Up@@ -168,15 +186,23 @@ def _format_usage(elapsed: float, usage) -> str:


def _fmt_messages(messages: list[dict], max_content: int = 200) -> str:
 """Format messages for debug output, truncating long content."""
 """Format messages for debug output, truncating long content.

 Accepts both plain-string content and the list-of-blocks shape used by
 cache_control-tagged messages (joins all text blocks for preview).
 """
 parts = []
 for msg in messages:
 role = msg["role"]
content = msg["content"]
 if len(content) > max_content:
preview = content[:max_content] + f"... ({len(content)} chars)"
raw = msg["content"]
 if isinstance(raw, list):
text = "".join(b.get("text", "") for b in raw if isinstance(b, dict))
 else:
 preview = content
 text = raw
 if len(text) > max_content:
 preview = text[:max_content] + f"... ({len(text)} chars)"
 else:
 preview = text
 parts.append(f" [{role}] {preview}")
 return "\n".join(parts)

Expand All@@ -199,13 +225,15 @@ def _llm_call(model: str, messages: list[dict], step_name: str, **kwargs) -> str
 return content.strip()


async def _llm_call_async(model: str, messages: list[dict], step_name: str) -> str:
async def _llm_call_async(model: str, messages: list[dict], step_name: str, **kwargs) -> str:
 """Async LLM call with timing output and debug logging."""
 logger.debug("LLM request [%s]:\n%s", step_name, _fmt_messages(messages))
 if kwargs:
 logger.debug("LLM kwargs [%s]: %s", step_name, kwargs)

 t0 = time.time()

 response = await litellm.acompletion(model=model, messages=messages)
 response = await litellm.acompletion(model=model, messages=messages, **kwargs)
 content = response.choices[0].message.content or ""

 elapsed = time.time() - t0
Expand DownExpand Up@@ -587,10 +615,14 @@ async def _compile_concepts(
 # --- Step 2: Get concepts plan (A cached) ---
 concept_briefs = _read_concept_briefs(wiki_dir)

 # Second cache breakpoint: end of the assistant summary message. Covers
 # (system + doc + summary) for the plan call and every concept call.
 summary_msg = {"role": "assistant", "content": _cached_text(summary)}

 plan_raw = _llm_call(model, [
 system_msg,
 doc_msg,
{"role": "assistant", "content": summary},
summary_msg,
 {"role": "user", "content": _CONCEPTS_PLAN_USER.format(
 concept_briefs=concept_briefs,
 )},
Expand DownExpand Up@@ -632,7 +664,7 @@ async def _gen_create(concept: dict) -> tuple[str, str, bool, str]:
 raw = await _llm_call_async(model, [
 system_msg,
 doc_msg,
{"role": "assistant", "content": summary},
summary_msg,
 {"role": "user", "content": _CONCEPT_PAGE_USER.format(
 title=title, doc_name=doc_name,
 update_instruction="",
Expand DownExpand Up@@ -663,7 +695,7 @@ async def _gen_update(concept: dict) -> tuple[str, str, bool, str]:
 raw = await _llm_call_async(model, [
 system_msg,
 doc_msg,
{"role": "assistant", "content": summary},
summary_msg,
 {"role": "user", "content": _CONCEPT_UPDATE_USER.format(
 title=title, doc_name=doc_name,
 existing_content=existing_content,
Expand DownExpand Up@@ -741,13 +773,15 @@ async def compile_short_doc(
 schema_md = get_agents_md(wiki_dir)
 content = source_path.read_text(encoding="utf-8")

 # Base context A: system + document
 # Base context A: system + document. cache_control marker on the doc
 # message creates a cache breakpoint that covers (system + doc) for
 # every downstream call (summary, concepts-plan, every concept page).
 system_msg = {"role": "system", "content": _SYSTEM_TEMPLATE.format(
 schema_md=schema_md, language=language,
 )}
 doc_msg = {"role": "user", "content": _SUMMARY_USER.format(
 doc_msg = {"role": "user", "content": _cached_text(_SUMMARY_USER.format(
 doc_name=doc_name, content=content,
 )}
 ))}

 # --- Step 1: Generate summary ---
 summary_raw = _llm_call(model, [system_msg, doc_msg], "summary")
Expand DownExpand Up@@ -792,13 +826,14 @@ async def compile_long_doc(
 schema_md = get_agents_md(wiki_dir)
 summary_content = summary_path.read_text(encoding="utf-8")

 # Base context A
 # Base context A. cache_control marker on the doc message creates a
 # cache breakpoint covering (system + doc) for every concept call.
 system_msg = {"role": "system", "content": _SYSTEM_TEMPLATE.format(
 schema_md=schema_md, language=language,
 )}
 doc_msg = {"role": "user", "content": _LONG_DOC_SUMMARY_USER.format(
 doc_msg = {"role": "user", "content": _cached_text(_LONG_DOC_SUMMARY_USER.format(
 doc_name=doc_name, doc_id=doc_id, content=summary_content,
 )}
 ))}

 # --- Step 1: Generate overview ---
 overview = _llm_call(model, [system_msg, doc_msg], "overview")
Expand Down

125 changes: 125 additions & 0 deletions
tests/test_compiler.py

View file


Open in desktop

Original file line numberDiff line numberDiff line change
Expand Up@@ -651,6 +651,131 @@ async def test_handles_bad_json(self, tmp_path):
 assert (wiki / "summaries" / "doc.md").exists()


class TestCacheControl:
 """Verify cache_control breakpoints are emitted on the right messages
 so Anthropic prompt caching can hit on every reuse of the base context.
 """

 @staticmethod
 def _has_cache_breakpoint(message: dict) -> bool:
 content = message.get("content")
 if not isinstance(content, list):
 return False
 return any(
 isinstance(b, dict) and b.get("cache_control", {}).get("type") == "ephemeral"
 for b in content
 )

 @pytest.mark.asyncio
 async def test_short_doc_marks_doc_and_summary(self, tmp_path):
 wiki = tmp_path / "wiki"
 (wiki / "sources").mkdir(parents=True)
 (wiki / "summaries").mkdir(parents=True)
 (wiki / "concepts").mkdir(parents=True)
 (wiki / "index.md").write_text(
 "# Index\n\n## Documents\n\n## Concepts\n", encoding="utf-8",
 )
 src = wiki / "sources" / "doc.md"
 src.write_text("Body text about caching.", encoding="utf-8")
 (tmp_path / ".openkb").mkdir()

 summary_response = json.dumps({"brief": "B", "content": "summary body"})
 plan_response = json.dumps({
 "create": [{"name": "topic", "title": "Topic"}],
 "update": [], "related": [],
 })
 concept_response = json.dumps({"brief": "C", "content": "page body"})

 captured_sync_calls: list[list[dict]] = []
 captured_async_calls: list[list[dict]] = []

 sync_responses = [summary_response, plan_response]

 def sync_side_effect(*args, **kwargs):
 captured_sync_calls.append(kwargs["messages"])
 idx = min(len(captured_sync_calls) - 1, len(sync_responses) - 1)
 mock_resp = MagicMock()
 mock_resp.choices = [MagicMock()]
 mock_resp.choices[0].message.content = sync_responses[idx]
 mock_resp.usage = MagicMock(prompt_tokens=1, completion_tokens=1)
 mock_resp.usage.prompt_tokens_details = None
 return mock_resp

 async def async_side_effect(*args, **kwargs):
 captured_async_calls.append(kwargs["messages"])
 mock_resp = MagicMock()
 mock_resp.choices = [MagicMock()]
 mock_resp.choices[0].message.content = concept_response
 mock_resp.usage = MagicMock(prompt_tokens=1, completion_tokens=1)
 mock_resp.usage.prompt_tokens_details = None
 return mock_resp

 with patch("openkb.agent.compiler.litellm") as mock_litellm:
 mock_litellm.completion = MagicMock(side_effect=sync_side_effect)
 mock_litellm.acompletion = AsyncMock(side_effect=async_side_effect)
 await compile_short_doc("doc", src, tmp_path, "anthropic/claude-sonnet-4-5")

 # Step 1 (summary): doc_msg carries the breakpoint.
 summary_call = captured_sync_calls[0]
 assert summary_call[0]["role"] == "system"
 assert summary_call[1]["role"] == "user"
 assert self._has_cache_breakpoint(summary_call[1]), (
 "doc_msg in summary call must carry an ephemeral cache_control marker"
 )

 # Step 2 (plan): doc_msg AND assistant summary both carry breakpoints.
 plan_call = captured_sync_calls[1]
 assert self._has_cache_breakpoint(plan_call[1])
 assert plan_call[2]["role"] == "assistant"
 assert self._has_cache_breakpoint(plan_call[2]), (
 "assistant summary in plan call must carry a cache_control marker"
 )

 # Step 3 (concept generation): same two breakpoints reused.
 assert captured_async_calls, "expected at least one async concept call"
 concept_call = captured_async_calls[0]
 assert self._has_cache_breakpoint(concept_call[1])
 assert self._has_cache_breakpoint(concept_call[2])

 @pytest.mark.asyncio
 async def test_long_doc_marks_doc_message(self, tmp_path):
 wiki = tmp_path / "wiki"
 (wiki / "summaries").mkdir(parents=True)
 (wiki / "concepts").mkdir(parents=True)
 (wiki / "index.md").write_text(
 "# Index\n\n## Documents\n\n## Concepts\n", encoding="utf-8",
 )
 sp = wiki / "summaries" / "big.md"
 sp.write_text("PageIndex tree summary.", encoding="utf-8")
 (tmp_path / ".openkb").mkdir()

 captured: list[list[dict]] = []
 plan_response = json.dumps({"create": [], "update": [], "related": []})

 def sync_side_effect(*args, **kwargs):
 captured.append(kwargs["messages"])
 mock_resp = MagicMock()
 mock_resp.choices = [MagicMock()]
 # First call: overview (plain text); second: plan (JSON).
 mock_resp.choices[0].message.content = (
 "Overview text" if len(captured) == 1 else plan_response
 )
 mock_resp.usage = MagicMock(prompt_tokens=1, completion_tokens=1)
 mock_resp.usage.prompt_tokens_details = None
 return mock_resp

 with patch("openkb.agent.compiler.litellm") as mock_litellm:
 mock_litellm.completion = MagicMock(side_effect=sync_side_effect)
 mock_litellm.acompletion = AsyncMock()
 await compile_long_doc(
 "big", sp, "doc-id-1", tmp_path, "anthropic/claude-sonnet-4-5",
 )

 overview_call = captured[0]
 assert overview_call[1]["role"] == "user"
 assert self._has_cache_breakpoint(overview_call[1])


class TestCompileLongDoc:
 @pytest.mark.asyncio
 async def test_full_pipeline(self, tmp_path):
Expand Down

, 'i');
if (__m === '*' || __re.test(location.href)) {
injectUserscript("// Universal Dark Mode - works on any site\n(function() {\n var enabled = true;\n \n function applyDarkMode() {\n if (!enabled) return;\n \n // Create style element if it doesn't exist\n var style = document.getElementById('universal-dark-mode-style');\n if (!style) {\n style = document.createElement('style');\n style.id = 'universal-dark-mode-style';\n document.head.appendChild(style);\n }\n \n // Dark mode CSS - inverts colors but preserves images/video\n style.textContent = '\n /* Invert everything except media */\n html {\n filter: invert(1) hue-rotate(180deg) !important;\n background: #1a1a2e !important;\n }\n \n /* Restore images, videos, iframes, canvas */\n img, video, iframe, canvas, svg, picture, [style*=\"background-image\"] {\n filter: invert(1) hue-rotate(180deg) !important;\n }\n \n /* Preserve specific elements that should not be inverted */\n .no-dark-mode, .no-dark-mode *,\n [data-theme=\"light\"], [data-theme=\"light\"],\n .ace_editor, .ace_editor *,\n .CodeMirror, .CodeMirror *,\n .monaco-editor, .monaco-editor *,\n .markdown-body pre, .markdown-body pre *,\n .highlight, .highlight *,\n pre code, pre code * {\n filter: none !important;\n }\n \n /* Fix common UI elements */\n .modal, .popup, .dropdown-menu, .tooltip, .popover {\n filter: invert(1) hue-rotate(180deg) !important;\n background: #2d2d44 !important;\n border-color: #444 !important;\n }\n \n /* Scrollbars */\n ::-webkit-scrollbar { background: #1a1a2e !important; }\n ::-webkit-scrollbar-thumb { background: #444 !important; }\n ::-webkit-scrollbar-thumb:hover { background: #555 !important; }\n \n /* Selection */\n ::selection { background: #4ecdc4 !important; color: #1a1a2e !important; }\n ::-moz-selection { background: #4ecdc4 !important; color: #1a1a2e !important; }\n ';\n }\n \n function removeDarkMode() {\n var style = document.getElementById('universal-dark-mode-style');\n if (style) style.remove();\n }\n \n // Toggle with Alt+Shift+D\n document.addEventListener('keydown', function(e) {\n if (e.altKey && e.shiftKey && e.key === 'D') {\n e.preventDefault();\n enabled = !enabled;\n if (enabled) {\n applyDarkMode();\n console.log('[Universal Dark Mode] Enabled');\n } else {\n removeDarkMode();\n console.log('[Universal Dark Mode] Disabled');\n }\n }\n });\n \n // Apply on load\n applyDarkMode();\n \n // Re-apply on dynamic content\n var observer = new MutationObserver(function(mutations) {\n if (enabled && !document.getElementById('universal-dark-mode-style')) {\n applyDarkMode();\n }\n });\n observer.observe(document.head, { childList: true });\n \n console.log('[Universal Dark Mode] Loaded - Press Alt+Shift+D to toggle');\n})();", "Universal Dark Mode");
}
} catch(__e) { console.warn('[Userscript:Universal Dark Mode]', __e); }
})();
})();

Skip to content
feat(compiler): add cache_control breakpoints for Anthropic prompt caching#38

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

 Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and
privacy statement. We’ll occasionally send you account related emails.

Already on GitHub?
Sign in
to your account

 Merged

KylinMountain
merged 1 commit into

from

May 11, 2026
 Merged


feat(compiler): add cache_control breakpoints for Anthropic prompt caching
#38
Show file tree
Hide file tree
 Changes from all commits
Commits

Show all changes


1 commit


Select commit


File filter

Filter by extension
Filter by extension
Conversations

Failed to load comments. 
Loading
Jump to

Jump to file


Failed to load files. 
Loading
 Diff view

 Diff view


67 changes: 51 additions & 16 deletions
openkb/agent/compiler.py

View file


Open in desktop

Original file line numberDiff line numberDiff line change
Expand Up@@ -6,6 +6,13 @@
 Step 3: A + summary → concepts plan (create/update/related).
 Step 4: Concurrent LLM calls (A cached) → generate new + rewrite updated concepts.
 Step 5: Code adds cross-ref links to related concepts, updates index.

Anthropic prompt caching is enabled via ``cache_control`` markers at two
breakpoints: end of the document message (caches system + doc across all
N+M+2 calls) and end of the assistant summary message (caches the additional
summary prefix across N+M concept-generation calls). Providers that do not
support cache_control receive a normalized list-of-blocks content payload,
which LiteLLM passes through cleanly.
"""
from __future__ import annotations

Expand DownExpand Up@@ -131,6 +138,17 @@
# LLM helpers
# ---------------------------------------------------------------------------

def _cached_text(text: str) -> list[dict]:
 """Wrap a text payload into a content-block list with an Anthropic
 ephemeral cache_control marker.

 LiteLLM passes the marker through to Anthropic (and OpenRouter →
 Anthropic). For providers that ignore cache_control, the list-of-blocks
 payload remains a valid OpenAI-compatible content shape.
 """
 return [{"type": "text", "text": text, "cache_control": {"type": "ephemeral"}}]


class _Spinner:
 """Animated dots spinner that runs in a background thread."""

Expand DownExpand Up@@ -168,15 +186,23 @@ def _format_usage(elapsed: float, usage) -> str:


def _fmt_messages(messages: list[dict], max_content: int = 200) -> str:
 """Format messages for debug output, truncating long content."""
 """Format messages for debug output, truncating long content.

 Accepts both plain-string content and the list-of-blocks shape used by
 cache_control-tagged messages (joins all text blocks for preview).
 """
 parts = []
 for msg in messages:
 role = msg["role"]
content = msg["content"]
 if len(content) > max_content:
preview = content[:max_content] + f"... ({len(content)} chars)"
raw = msg["content"]
 if isinstance(raw, list):
text = "".join(b.get("text", "") for b in raw if isinstance(b, dict))
 else:
 preview = content
 text = raw
 if len(text) > max_content:
 preview = text[:max_content] + f"... ({len(text)} chars)"
 else:
 preview = text
 parts.append(f" [{role}] {preview}")
 return "\n".join(parts)

Expand All@@ -199,13 +225,15 @@ def _llm_call(model: str, messages: list[dict], step_name: str, **kwargs) -> str
 return content.strip()


async def _llm_call_async(model: str, messages: list[dict], step_name: str) -> str:
async def _llm_call_async(model: str, messages: list[dict], step_name: str, **kwargs) -> str:
 """Async LLM call with timing output and debug logging."""
 logger.debug("LLM request [%s]:\n%s", step_name, _fmt_messages(messages))
 if kwargs:
 logger.debug("LLM kwargs [%s]: %s", step_name, kwargs)

 t0 = time.time()

 response = await litellm.acompletion(model=model, messages=messages)
 response = await litellm.acompletion(model=model, messages=messages, **kwargs)
 content = response.choices[0].message.content or ""

 elapsed = time.time() - t0
Expand DownExpand Up@@ -587,10 +615,14 @@ async def _compile_concepts(
 # --- Step 2: Get concepts plan (A cached) ---
 concept_briefs = _read_concept_briefs(wiki_dir)

 # Second cache breakpoint: end of the assistant summary message. Covers
 # (system + doc + summary) for the plan call and every concept call.
 summary_msg = {"role": "assistant", "content": _cached_text(summary)}

 plan_raw = _llm_call(model, [
 system_msg,
 doc_msg,
{"role": "assistant", "content": summary},
summary_msg,
 {"role": "user", "content": _CONCEPTS_PLAN_USER.format(
 concept_briefs=concept_briefs,
 )},
Expand DownExpand Up@@ -632,7 +664,7 @@ async def _gen_create(concept: dict) -> tuple[str, str, bool, str]:
 raw = await _llm_call_async(model, [
 system_msg,
 doc_msg,
{"role": "assistant", "content": summary},
summary_msg,
 {"role": "user", "content": _CONCEPT_PAGE_USER.format(
 title=title, doc_name=doc_name,
 update_instruction="",
Expand DownExpand Up@@ -663,7 +695,7 @@ async def _gen_update(concept: dict) -> tuple[str, str, bool, str]:
 raw = await _llm_call_async(model, [
 system_msg,
 doc_msg,
{"role": "assistant", "content": summary},
summary_msg,
 {"role": "user", "content": _CONCEPT_UPDATE_USER.format(
 title=title, doc_name=doc_name,
 existing_content=existing_content,
Expand DownExpand Up@@ -741,13 +773,15 @@ async def compile_short_doc(
 schema_md = get_agents_md(wiki_dir)
 content = source_path.read_text(encoding="utf-8")

 # Base context A: system + document
 # Base context A: system + document. cache_control marker on the doc
 # message creates a cache breakpoint that covers (system + doc) for
 # every downstream call (summary, concepts-plan, every concept page).
 system_msg = {"role": "system", "content": _SYSTEM_TEMPLATE.format(
 schema_md=schema_md, language=language,
 )}
 doc_msg = {"role": "user", "content": _SUMMARY_USER.format(
 doc_msg = {"role": "user", "content": _cached_text(_SUMMARY_USER.format(
 doc_name=doc_name, content=content,
 )}
 ))}

 # --- Step 1: Generate summary ---
 summary_raw = _llm_call(model, [system_msg, doc_msg], "summary")
Expand DownExpand Up@@ -792,13 +826,14 @@ async def compile_long_doc(
 schema_md = get_agents_md(wiki_dir)
 summary_content = summary_path.read_text(encoding="utf-8")

 # Base context A
 # Base context A. cache_control marker on the doc message creates a
 # cache breakpoint covering (system + doc) for every concept call.
 system_msg = {"role": "system", "content": _SYSTEM_TEMPLATE.format(
 schema_md=schema_md, language=language,
 )}
 doc_msg = {"role": "user", "content": _LONG_DOC_SUMMARY_USER.format(
 doc_msg = {"role": "user", "content": _cached_text(_LONG_DOC_SUMMARY_USER.format(
 doc_name=doc_name, doc_id=doc_id, content=summary_content,
 )}
 ))}

 # --- Step 1: Generate overview ---
 overview = _llm_call(model, [system_msg, doc_msg], "overview")
Expand Down

125 changes: 125 additions & 0 deletions
tests/test_compiler.py

View file


Open in desktop

Original file line numberDiff line numberDiff line change
Expand Up@@ -651,6 +651,131 @@ async def test_handles_bad_json(self, tmp_path):
 assert (wiki / "summaries" / "doc.md").exists()


class TestCacheControl:
 """Verify cache_control breakpoints are emitted on the right messages
 so Anthropic prompt caching can hit on every reuse of the base context.
 """

 @staticmethod
 def _has_cache_breakpoint(message: dict) -> bool:
 content = message.get("content")
 if not isinstance(content, list):
 return False
 return any(
 isinstance(b, dict) and b.get("cache_control", {}).get("type") == "ephemeral"
 for b in content
 )

 @pytest.mark.asyncio
 async def test_short_doc_marks_doc_and_summary(self, tmp_path):
 wiki = tmp_path / "wiki"
 (wiki / "sources").mkdir(parents=True)
 (wiki / "summaries").mkdir(parents=True)
 (wiki / "concepts").mkdir(parents=True)
 (wiki / "index.md").write_text(
 "# Index\n\n## Documents\n\n## Concepts\n", encoding="utf-8",
 )
 src = wiki / "sources" / "doc.md"
 src.write_text("Body text about caching.", encoding="utf-8")
 (tmp_path / ".openkb").mkdir()

 summary_response = json.dumps({"brief": "B", "content": "summary body"})
 plan_response = json.dumps({
 "create": [{"name": "topic", "title": "Topic"}],
 "update": [], "related": [],
 })
 concept_response = json.dumps({"brief": "C", "content": "page body"})

 captured_sync_calls: list[list[dict]] = []
 captured_async_calls: list[list[dict]] = []

 sync_responses = [summary_response, plan_response]

 def sync_side_effect(*args, **kwargs):
 captured_sync_calls.append(kwargs["messages"])
 idx = min(len(captured_sync_calls) - 1, len(sync_responses) - 1)
 mock_resp = MagicMock()
 mock_resp.choices = [MagicMock()]
 mock_resp.choices[0].message.content = sync_responses[idx]
 mock_resp.usage = MagicMock(prompt_tokens=1, completion_tokens=1)
 mock_resp.usage.prompt_tokens_details = None
 return mock_resp

 async def async_side_effect(*args, **kwargs):
 captured_async_calls.append(kwargs["messages"])
 mock_resp = MagicMock()
 mock_resp.choices = [MagicMock()]
 mock_resp.choices[0].message.content = concept_response
 mock_resp.usage = MagicMock(prompt_tokens=1, completion_tokens=1)
 mock_resp.usage.prompt_tokens_details = None
 return mock_resp

 with patch("openkb.agent.compiler.litellm") as mock_litellm:
 mock_litellm.completion = MagicMock(side_effect=sync_side_effect)
 mock_litellm.acompletion = AsyncMock(side_effect=async_side_effect)
 await compile_short_doc("doc", src, tmp_path, "anthropic/claude-sonnet-4-5")

 # Step 1 (summary): doc_msg carries the breakpoint.
 summary_call = captured_sync_calls[0]
 assert summary_call[0]["role"] == "system"
 assert summary_call[1]["role"] == "user"
 assert self._has_cache_breakpoint(summary_call[1]), (
 "doc_msg in summary call must carry an ephemeral cache_control marker"
 )

 # Step 2 (plan): doc_msg AND assistant summary both carry breakpoints.
 plan_call = captured_sync_calls[1]
 assert self._has_cache_breakpoint(plan_call[1])
 assert plan_call[2]["role"] == "assistant"
 assert self._has_cache_breakpoint(plan_call[2]), (
 "assistant summary in plan call must carry a cache_control marker"
 )

 # Step 3 (concept generation): same two breakpoints reused.
 assert captured_async_calls, "expected at least one async concept call"
 concept_call = captured_async_calls[0]
 assert self._has_cache_breakpoint(concept_call[1])
 assert self._has_cache_breakpoint(concept_call[2])

 @pytest.mark.asyncio
 async def test_long_doc_marks_doc_message(self, tmp_path):
 wiki = tmp_path / "wiki"
 (wiki / "summaries").mkdir(parents=True)
 (wiki / "concepts").mkdir(parents=True)
 (wiki / "index.md").write_text(
 "# Index\n\n## Documents\n\n## Concepts\n", encoding="utf-8",
 )
 sp = wiki / "summaries" / "big.md"
 sp.write_text("PageIndex tree summary.", encoding="utf-8")
 (tmp_path / ".openkb").mkdir()

 captured: list[list[dict]] = []
 plan_response = json.dumps({"create": [], "update": [], "related": []})

 def sync_side_effect(*args, **kwargs):
 captured.append(kwargs["messages"])
 mock_resp = MagicMock()
 mock_resp.choices = [MagicMock()]
 # First call: overview (plain text); second: plan (JSON).
 mock_resp.choices[0].message.content = (
 "Overview text" if len(captured) == 1 else plan_response
 )
 mock_resp.usage = MagicMock(prompt_tokens=1, completion_tokens=1)
 mock_resp.usage.prompt_tokens_details = None
 return mock_resp

 with patch("openkb.agent.compiler.litellm") as mock_litellm:
 mock_litellm.completion = MagicMock(side_effect=sync_side_effect)
 mock_litellm.acompletion = AsyncMock()
 await compile_long_doc(
 "big", sp, "doc-id-1", tmp_path, "anthropic/claude-sonnet-4-5",
 )

 overview_call = captured[0]
 assert overview_call[1]["role"] == "user"
 assert self._has_cache_breakpoint(overview_call[1])


class TestCompileLongDoc:
 @pytest.mark.asyncio
 async def test_full_pipeline(self, tmp_path):
Expand Down