feat: add memory/document management to Agno, Mastra, and Raycast plugins

packages/plugin-agno/.gitignore

@@ -0,0 +1,23 @@
+node_modules/
+dist/
+package-lock.json
+npm-debug.log*
+yarn.lock
+.pnpm-debug.log*
+
+*.tsbuildinfo
+
+.venv/
+__pycache__/
+*.py[cod]
+*.pyo
+*.pyd
+*.egg-info/
+.eggs/
+
+.DS_Store
+.vscode/
+.idea/
+*.swp
+*.swo
+

packages/plugin-agno/README.md

@@ -41,14 +41,29 @@ agent.print_response("What theme do I prefer?", stream=True)
 
 ## Available tools
 
-The `NeocortexTools` toolkit exposes three tools to the agent:
+The `NeocortexTools` toolkit exposes tools aligned with the TinyHumans/Neocortex SDK endpoints:
 
 
 | Tool | Description |
 | --------------- | --------------------------------------------------------------------- |
 | `save_memory` | Save or update a memory (key, content, namespace, optional metadata). |
 | `recall_memory` | Recall relevant memories for a natural-language query in a namespace. |
 | `delete_memory` | Delete one or more memories by key/keys or delete all in a namespace. |
+| `sync_memory` | Sync OpenClaw memory files (workspace/agent + file objects). |
+| `insert_document` | Insert a single memory document (title/content/namespace). |
+| `insert_documents_batch` | Insert multiple documents in one call. |
+| `list_documents` | List documents in a namespace. |
+| `get_document` | Get a specific document by `document_id`. |
+| `delete_document` | Delete a specific document by `document_id`. |
+| `query_memory_context` | Query mirrored memory context (`/v1/memory/queries`). |
+| `chat_memory_context` | Chat with memory context (`/v1/memory/conversations`). |
+| `record_interactions` | Record interaction signals (`/v1/memory/interactions`). |
+| `recall_thoughts` | Generate reflective thoughts (`/v1/memory/memories/thoughts`). |
+| `chat_memory` | Chat with memory cache (`/v1/memory/chat`). |
+| `interact_memory` | Record entity interactions (`/v1/memory/interact`). |
+| `recall_memory_master` | Recall context from the master node (`/v1/memory/recall`). |
+| `recall_memories` | Recall memories from the Ebbinghaus bank (`/v1/memory/memories/recall`). |
+| `get_ingestion_job` | Check ingestion job status (`/v1/memory/ingestion/jobs/:jobId`). |
 
 
 Credentials (`token`, `model_id`, `base_url`) are set when constructing `NeocortexTools` and are **never** passed as tool arguments, so the LLM cannot see or override them.
@@ -60,7 +75,7 @@ Credentials (`token`, `model_id`, `base_url`) are set when constructing `Neocort
 
 ## Error handling
 
-On API failures, the underlying client raises `TinyHumanError`. You can catch it for logging or user-facing messages:
+On API failures, the underlying client raises `AlphahumanError`. You can catch it for logging or user-facing messages:
 
 ```python
 from neocortex_agno import NeocortexTools, AlphahumanError

packages/plugin-agno/example.py

@@ -5,6 +5,11 @@
  export ALPHAHUMAN_BASE_URL=""
  export OPENAI_API_KEY=""
  python example.py
+
+This example demonstrates both:
+- Saving/recalling simple memories (preferences)
+- Document + context workflows (insert/list/get documents, query/chat context,
+ record interactions, and recall thoughts)
 """
 
 import os
@@ -29,9 +34,16 @@ def main() -> None:
  )
  ],
  instructions=(
- "Use the memory tools to remember and recall user preferences and context. "
- "When the user tells you something to remember, use save_memory. "
- "When answering questions that might use stored context, use recall_memory first."
+ "Use the memory tools to remember and recall user preferences and context.\n"
+ "When the user tells you something to remember, use save_memory.\n"
+ "When answering questions that might use stored context, use recall_memory first.\n"
+ "If the user asks about documents or document-backed context, use:\n"
+ "- insert_document / insert_documents_batch\n"
+ "- list_documents / get_document\n"
+ "- query_memory_context (POST /v1/memory/queries)\n"
+ "- chat_memory_context (POST /v1/memory/conversations)\n"
+ "If the user asks to track signal-level memory, use record_interactions.\n"
+ "If the user asks for reflective/summary insights from memory, use recall_thoughts."
  ),
  markdown=True,
  )
@@ -47,6 +59,22 @@ def main() -> None:
  )
  agent.print_response("What theme do I prefer?", stream=True)
 
+ print()
+ print("Document + context workflow:")
+ agent.print_response(
+ "Create a document in namespace 'agno-docs' titled 'Alex Preferences'. "
+ "Store the content: 'Alex prefers dark mode and wants succinct answers.'. "
+ "Next, query_memory_context in 'agno-docs' for: 'What does Alex prefer?' "
+ "and use that output to answer. "
+ "Then call chat_memory_context with messages=[{'role':'user','content':'What does Alex prefer?'}] "
+ "using the same namespace 'agno-docs' context. "
+ "After that, call record_interactions in 'agno-docs' with "
+ "entity_names=['ENTITY-AGNO-A','ENTITY-AGNO-B'] and interaction_level='engage'. "
+ "Finally, call recall_thoughts for 'agno-docs' with max_chunks=5. "
+ "Return a short summary of each step's outcome.",
+ stream=True,
+ )
+
 
 if __name__ == "__main__":
  main()