fix(core): expand model.provider runtime JSON schema enum to match Provider variants (#345 follow-up) by moonming · Pull Request #346 · api7/aisix

moonming · 2026-05-18T17:56:54Z

Summary

ai-gateway #345 (P2-A) added 11 long-tail OpenAI-adapter Provider::* variants + Hub registrations. The dashboard-side schema in schemas/resources/model.schema.json regenerated correctly. But there's a SEPARATELY-MAINTAINED hardcoded JSON schema in crates/aisix-core/src/models/schema.rs::model_schema() that backs aisix-etcd::loader's runtime validate_model check — and nobody updated it.

Result: cp-api admits a groq/mistral/etc. ProviderKey and Model (per #336 long-tail admission). The DP's aisix-etcd::loader reads the Model from etcd and validates against the OLD 6-value schema:

schema validation failed at `/provider`: "groq" is not one of ["openai","anthropic","google","deepseek","cohere","jina"]

Model rejected. Never appears in /v1/models. Customer's chat 404s on the alias.

Surfaced by AISIX-Cloud PR #366 e2e — first iteration of the D3.2 long-tail matrix. The groq Model creation succeeded at cp-api but the DP never saw it. Logs showed the schema rejection.

Fix

crates/aisix-core/src/models/schema.rs:120 — extend the provider enum from 6 to 17 values to match every Provider::as_str() output post-#345.

The existing negative-test model_unknown_provider_value_fails used "mistral" as the "unknown value" sentinel. Since mistral is now valid, renamed the fixture string to "this-is-not-a-provider-id" so the negative-test still pins the rejection class.

Why a separate schema even exists

aisix-etcd::loader::validate_and_parse runs schema validation BEFORE serde deserialization to fail-loud-and-skip on malformed entries (rather than crashing the DP). The schema is the gate; serde is the parser. Both must agree on the enum.

The dashboard-facing JSON schemas in schemas/resources/*.schema.json are GENERATED from the Rust types via cargo run --bin dump-schema (drift CI check enforces). The runtime validator's schema in schema.rs is hand-written and not gated by the drift check — a structural gap that this PR addresses with a behavioral fix only; a future refactor could fold the two sources together.

Test plan

cargo test -p aisix-core --lib schema — 52 tests pass
clippy clean
fmt clean
CI green
Audit (CLAUDE.md §7)
After merge + docker rebuild, AISIX-Cloud chore(fmt): restore rustfmt on aisix-etcd loader test (post-#362) #366 D3.2 batch 1 e2e exercises the unblocked path

References (per CLAUDE.md §7)

ai-gateway#345 (merged) — Provider enum + Hub register
aisix-etcd loader.rs:262 — the schema-rejection log path that surfaced the gap on PR chore(fmt): restore rustfmt on aisix-etcd loader test (post-#362) #366 CI

Refs api7/AISIX-Cloud#366

Summary by CodeRabbit

Improvements
- Expanded model provider validation to accept an extended set of provider identifiers during schema validation.
Tests
- Updated test cases to reflect the expanded provider validation rules, ensuring invalid providers continue to be properly rejected.

…ovider variants (#345 follow-up) ai-gateway #345 added 11 long-tail OpenAI-adapter Provider variants (groq, mistral, togetherai, fireworks-ai, perplexity, moonshotai, alibaba, zhipuai, baseten, huggingface, cerebras) to the Rust enum + Hub registrations + cp-api admission. The hardcoded JSON schema backing `aisix-etcd::loader`'s `validate_model` runtime check at `crates/aisix-core/src/models/schema.rs:120`, however, retained the original 6-value allowlist. The dashboard-side schema in `schemas/resources/model.schema.json` was regenerated correctly by #345's `dump-schema` step — that file IS now in sync with the enum. But the runtime validator uses a SEPARATELY-MAINTAINED hardcoded schema at line 112-200 of schema.rs that nobody updated. cp-api admission ✓, DP serde ✓, runtime schema validator ✗ — a customer's groq Model resource gets rejected at the etcd loader before deserialization with `schema validation failed at \`/provider\`: "groq" is not one of ["openai","anthropic","google","deepseek","cohere","jina"]`. This silently dropped EVERY long-tail Provider Model resource the dashboard wrote. The model never appears in `/v1/models` on the DP, and customer chat calls 404 on the alias. Surfaced by AISIX-Cloud PR #366 e2e (D3.2 batch 1) — the groq Model creation succeeds at cp-api but the DP never sees it. Logs showed the schema validation rejection. Fix `crates/aisix-core/src/models/schema.rs:120` — extend the `provider` enum to the full 17 values matching `Provider::as_str()` output across every variant: ```rust "provider": { "type": "string", "enum": [ "openai","anthropic","google","deepseek","cohere","jina", "groq","mistral","togetherai","fireworks-ai","perplexity", "moonshotai","alibaba","zhipuai","baseten","huggingface","cerebras" ] }, ``` The existing negative-test `model_unknown_provider_value_fails` used `"mistral"` as the "unknown value" sentinel; since mistral is now valid, the test would have stopped catching the rejection path. Renamed the test fixture value to `"this-is-not-a-provider-id"` so the negative-test still pins the rejection class. All 52 aisix-core schema tests pass; clippy clean; fmt clean. References (per CLAUDE.md §7) - ai-gateway#345 (merged) — Provider enum + Hub register additions - aisix-etcd loader.rs:262 — schema-rejection log path that surfaced the gap

coderabbitai · 2026-05-18T17:57:10Z

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Free

Run ID: 6608819e-114c-4760-8c03-7fa8bfd7b4c6

📥 Commits

Reviewing files that changed from the base of the PR and between 8d88b41 and ab76aa5.

📒 Files selected for processing (3)

crates/aisix-admin/src/lib.rs
crates/aisix-etcd/src/loader.rs
crates/aisix-etcd/src/supervisor.rs

✅ Files skipped from review due to trivial changes (1)

crates/aisix-etcd/src/supervisor.rs

📝 Walkthrough

Walkthrough

The PR expands the model JSON schema's provider field enum to accept a larger set of provider identifiers and updates tests/fixtures that previously used mistral to use a clearly invalid provider string so schema-rejection tests still fail.

Changes

Provider Field Allowlist Expansion

Layer / File(s)	Summary
Provider field schema and core validation test `crates/aisix-core/src/models/schema.rs`	The `provider` field enum in the model schema is expanded to accept a larger allowlist. The `model_unknown_provider_value_fails` test is updated to use an explicitly invalid provider string so the validation rejection path is still exercised.
Update tests and fixtures across crates `crates/aisix-admin/src/lib.rs`, `crates/aisix-etcd/src/loader.rs`, `crates/aisix-etcd/src/supervisor.rs`	Tests and JSON fixtures that previously used `provider: "mistral"` were changed to use `provider: "this-is-not-a-provider-id"` (or similar) to keep negative schema-validation tests failing under the expanded provider allowlist.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Note

🎁 Summarized by CodeRabbit Free

Your organization has reached its limit of developer seats under the Pro Plan. For new users, CodeRabbit will generate a high-level summary and a walkthrough for each pull request. For a comprehensive line-by-line review, please add seats to your subscription by visiting https://app.coderabbit.ai/login.If you believe this is a mistake and have available seats, please assign one to the pull request author through the subscription management page using the link above.

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

…rovider sentinel The negative-test fixtures across `aisix-admin` and `aisix-etcd` used `"mistral"` as the "unknown provider" sentinel — a vestige of the pre-#345 6-value allowlist. With ai-gateway#345 first-classing 11 long-tail variants (including mistral), those tests started asserting against a now-VALID provider and FAILED. Replaced every fixture occurrence with `"this-is-not-a-provider-id"`, which is intentionally NOT in the post-#345 17-value enum, keeping the schema-rejection class pinned without churn-tracking the live provider list. Files touched (post-#345 surgical sentinel rotation): - `crates/aisix-admin/src/lib.rs::create_model_with_invalid_provider_prefix_is_400_schema_error` - `crates/aisix-etcd/src/loader.rs::tests` (2 sites) - `crates/aisix-etcd/src/supervisor.rs::tests` (1 site, drives ~6 transitive supervisor test failures via shared fixture) All workspace tests green; clippy clean; fmt clean. Refs ai-gateway#345 ai-gateway#346

Integrate origin/main (commit 2c1d485 = post-PR-#326 / #348 plus #330 / #341 / #343 / #345 / #346) into this branch via `git merge --squash` to clear PR #344's lingering `mergeable: dirty` state. Conflict on `docs/quickstart/self-hosted.md` was a 3-way-merge-base artifact: base (3596c0a) read `- a reachable etcd instance`, main changed `a` → `A` (via #326), this branch additionally inserted the glossary link. Both changes are wanted; resolution per Umar's approved plan was `git checkout --ours`, which preserves the branch's self-hosted.md state (already integrates capital A + glossary link + first-time-build paragraph + keep-running framing). Other 4 overlapping doc files auto-merged cleanly (`bootstrap-config.md`, `core-concepts.md`, `first-model-first-key-first-request.md`, `openai-sdk.md`). Code files all auto-merged cleanly. Additional Copilot review (post-`167196a` cycle) addressed: - `docs/index.md:7` — change link display text from `[data-plane]` to `[data plane]` to match the canonical glossary term. The URL anchor `#data-plane` stays kebab-case (matches the glossary heading's auto-anchor); only the display text changes. Comment id 3271145422. - `docs/quickstart/openai-sdk.md:43` — change `All three steps below` to `All commands below`. The Install-the-SDK section has two command blocks (mkdir+cd, npm install), not three; the prior wording originated from a mental model (mkdir, cd, install) that doesn't match the typographic count of code blocks under the heading. Comment id 3271145458. Copilot's third comment on `docs/overview/core-concepts.md` Observability Exporter wording (id 3271145444) auto-resolves via this merge — main's #326 rewrite supersedes the branch's pre-#326 wording at that location ("ships per-request span telemetry… OTLP/HTTP-compatible backend…" replaces "Use this concept when documenting…"). No separate edit needed; the merge IS the fix.

…ickstart-polish Resolve PR #344's lingering mergeable: dirty state by linking the branch history to origin/main (2c1d485 = post-#326 / #348 / #330 / #341 / #343 / #345 / #346). The squash-merge commit landed earlier (e2af197) integrated main's content into the branch tree but did not link the histories, so GitHub's mergeable computation still saw the 3-way-merge-base artifact conflict on docs/quickstart/self-hosted.md (a vs A + the glossary link / "In another terminal" vs "Keep the gateway running" framing). This explicit merge commit ties the branch to main's history. Self-hosted.md conflict resolved by taking OUR side — the branch's edits already contain main's substantive changes (capital A, first-time-build paragraph) plus this PR's additions (glossary link, keep-running framing, YOUR_ADMIN_KEY note, config.yaml location anchor). The auto-merge of first-model-first-key-first-request.md duplicated the :::warning callout that was already integrated via the squash commit; removed the duplicate.

moonming · 2026-05-18T17:56:54Z

Summary

ai-gateway #345 (P2-A) added 11 long-tail OpenAI-adapter Provider::* variants + Hub registrations. The dashboard-side schema in schemas/resources/model.schema.json regenerated correctly. But there's a SEPARATELY-MAINTAINED hardcoded JSON schema in crates/aisix-core/src/models/schema.rs::model_schema() that backs aisix-etcd::loader's runtime validate_model check — and nobody updated it.

Result: cp-api admits a groq/mistral/etc. ProviderKey and Model (per #336 long-tail admission). The DP's aisix-etcd::loader reads the Model from etcd and validates against the OLD 6-value schema:

schema validation failed at `/provider`: "groq" is not one of ["openai","anthropic","google","deepseek","cohere","jina"]

Model rejected. Never appears in /v1/models. Customer's chat 404s on the alias.

Surfaced by AISIX-Cloud PR #366 e2e — first iteration of the D3.2 long-tail matrix. The groq Model creation succeeded at cp-api but the DP never saw it. Logs showed the schema rejection.

Fix

crates/aisix-core/src/models/schema.rs:120 — extend the provider enum from 6 to 17 values to match every Provider::as_str() output post-#345.

The existing negative-test model_unknown_provider_value_fails used "mistral" as the "unknown value" sentinel. Since mistral is now valid, renamed the fixture string to "this-is-not-a-provider-id" so the negative-test still pins the rejection class.

Why a separate schema even exists

aisix-etcd::loader::validate_and_parse runs schema validation BEFORE serde deserialization to fail-loud-and-skip on malformed entries (rather than crashing the DP). The schema is the gate; serde is the parser. Both must agree on the enum.

The dashboard-facing JSON schemas in schemas/resources/*.schema.json are GENERATED from the Rust types via cargo run --bin dump-schema (drift CI check enforces). The runtime validator's schema in schema.rs is hand-written and not gated by the drift check — a structural gap that this PR addresses with a behavioral fix only; a future refactor could fold the two sources together.

Test plan

cargo test -p aisix-core --lib schema — 52 tests pass
clippy clean
fmt clean
CI green
Audit (CLAUDE.md §7)
After merge + docker rebuild, AISIX-Cloud chore(fmt): restore rustfmt on aisix-etcd loader test (post-#362) #366 D3.2 batch 1 e2e exercises the unblocked path

References (per CLAUDE.md §7)

ai-gateway#345 (merged) — Provider enum + Hub register
aisix-etcd loader.rs:262 — the schema-rejection log path that surfaced the gap on PR chore(fmt): restore rustfmt on aisix-etcd loader test (post-#362) #366 CI

Refs api7/AISIX-Cloud#366

Summary by CodeRabbit

Improvements
- Expanded model provider validation to accept an extended set of provider identifiers during schema validation.
Tests
- Updated test cases to reflect the expanded provider validation rules, ensuring invalid providers continue to be properly rejected.

…ovider variants (#345 follow-up) ai-gateway #345 added 11 long-tail OpenAI-adapter Provider variants (groq, mistral, togetherai, fireworks-ai, perplexity, moonshotai, alibaba, zhipuai, baseten, huggingface, cerebras) to the Rust enum + Hub registrations + cp-api admission. The hardcoded JSON schema backing `aisix-etcd::loader`'s `validate_model` runtime check at `crates/aisix-core/src/models/schema.rs:120`, however, retained the original 6-value allowlist. The dashboard-side schema in `schemas/resources/model.schema.json` was regenerated correctly by #345's `dump-schema` step — that file IS now in sync with the enum. But the runtime validator uses a SEPARATELY-MAINTAINED hardcoded schema at line 112-200 of schema.rs that nobody updated. cp-api admission ✓, DP serde ✓, runtime schema validator ✗ — a customer's groq Model resource gets rejected at the etcd loader before deserialization with `schema validation failed at \`/provider\`: "groq" is not one of ["openai","anthropic","google","deepseek","cohere","jina"]`. This silently dropped EVERY long-tail Provider Model resource the dashboard wrote. The model never appears in `/v1/models` on the DP, and customer chat calls 404 on the alias. Surfaced by AISIX-Cloud PR #366 e2e (D3.2 batch 1) — the groq Model creation succeeds at cp-api but the DP never sees it. Logs showed the schema validation rejection. Fix `crates/aisix-core/src/models/schema.rs:120` — extend the `provider` enum to the full 17 values matching `Provider::as_str()` output across every variant: ```rust "provider": { "type": "string", "enum": [ "openai","anthropic","google","deepseek","cohere","jina", "groq","mistral","togetherai","fireworks-ai","perplexity", "moonshotai","alibaba","zhipuai","baseten","huggingface","cerebras" ] }, ``` The existing negative-test `model_unknown_provider_value_fails` used `"mistral"` as the "unknown value" sentinel; since mistral is now valid, the test would have stopped catching the rejection path. Renamed the test fixture value to `"this-is-not-a-provider-id"` so the negative-test still pins the rejection class. All 52 aisix-core schema tests pass; clippy clean; fmt clean. References (per CLAUDE.md §7) - ai-gateway#345 (merged) — Provider enum + Hub register additions - aisix-etcd loader.rs:262 — schema-rejection log path that surfaced the gap

coderabbitai · 2026-05-18T17:57:10Z

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Free

Run ID: 6608819e-114c-4760-8c03-7fa8bfd7b4c6

📥 Commits

Reviewing files that changed from the base of the PR and between 8d88b41 and ab76aa5.

📒 Files selected for processing (3)

crates/aisix-admin/src/lib.rs
crates/aisix-etcd/src/loader.rs
crates/aisix-etcd/src/supervisor.rs

✅ Files skipped from review due to trivial changes (1)

crates/aisix-etcd/src/supervisor.rs

📝 Walkthrough

Walkthrough

The PR expands the model JSON schema's provider field enum to accept a larger set of provider identifiers and updates tests/fixtures that previously used mistral to use a clearly invalid provider string so schema-rejection tests still fail.

Changes

Provider Field Allowlist Expansion

Layer / File(s)	Summary
Provider field schema and core validation test `crates/aisix-core/src/models/schema.rs`	The `provider` field enum in the model schema is expanded to accept a larger allowlist. The `model_unknown_provider_value_fails` test is updated to use an explicitly invalid provider string so the validation rejection path is still exercised.
Update tests and fixtures across crates `crates/aisix-admin/src/lib.rs`, `crates/aisix-etcd/src/loader.rs`, `crates/aisix-etcd/src/supervisor.rs`	Tests and JSON fixtures that previously used `provider: "mistral"` were changed to use `provider: "this-is-not-a-provider-id"` (or similar) to keep negative schema-validation tests failing under the expanded provider allowlist.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Note

🎁 Summarized by CodeRabbit Free

Your organization has reached its limit of developer seats under the Pro Plan. For new users, CodeRabbit will generate a high-level summary and a walkthrough for each pull request. For a comprehensive line-by-line review, please add seats to your subscription by visiting https://app.coderabbit.ai/login.If you believe this is a mistake and have available seats, please assign one to the pull request author through the subscription management page using the link above.

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

…rovider sentinel The negative-test fixtures across `aisix-admin` and `aisix-etcd` used `"mistral"` as the "unknown provider" sentinel — a vestige of the pre-#345 6-value allowlist. With ai-gateway#345 first-classing 11 long-tail variants (including mistral), those tests started asserting against a now-VALID provider and FAILED. Replaced every fixture occurrence with `"this-is-not-a-provider-id"`, which is intentionally NOT in the post-#345 17-value enum, keeping the schema-rejection class pinned without churn-tracking the live provider list. Files touched (post-#345 surgical sentinel rotation): - `crates/aisix-admin/src/lib.rs::create_model_with_invalid_provider_prefix_is_400_schema_error` - `crates/aisix-etcd/src/loader.rs::tests` (2 sites) - `crates/aisix-etcd/src/supervisor.rs::tests` (1 site, drives ~6 transitive supervisor test failures via shared fixture) All workspace tests green; clippy clean; fmt clean. Refs ai-gateway#345 ai-gateway#346

Integrate origin/main (commit 2c1d485 = post-PR-#326 / #348 plus #330 / #341 / #343 / #345 / #346) into this branch via `git merge --squash` to clear PR #344's lingering `mergeable: dirty` state. Conflict on `docs/quickstart/self-hosted.md` was a 3-way-merge-base artifact: base (3596c0a) read `- a reachable etcd instance`, main changed `a` → `A` (via #326), this branch additionally inserted the glossary link. Both changes are wanted; resolution per Umar's approved plan was `git checkout --ours`, which preserves the branch's self-hosted.md state (already integrates capital A + glossary link + first-time-build paragraph + keep-running framing). Other 4 overlapping doc files auto-merged cleanly (`bootstrap-config.md`, `core-concepts.md`, `first-model-first-key-first-request.md`, `openai-sdk.md`). Code files all auto-merged cleanly. Additional Copilot review (post-`167196a` cycle) addressed: - `docs/index.md:7` — change link display text from `[data-plane]` to `[data plane]` to match the canonical glossary term. The URL anchor `#data-plane` stays kebab-case (matches the glossary heading's auto-anchor); only the display text changes. Comment id 3271145422. - `docs/quickstart/openai-sdk.md:43` — change `All three steps below` to `All commands below`. The Install-the-SDK section has two command blocks (mkdir+cd, npm install), not three; the prior wording originated from a mental model (mkdir, cd, install) that doesn't match the typographic count of code blocks under the heading. Comment id 3271145458. Copilot's third comment on `docs/overview/core-concepts.md` Observability Exporter wording (id 3271145444) auto-resolves via this merge — main's #326 rewrite supersedes the branch's pre-#326 wording at that location ("ships per-request span telemetry… OTLP/HTTP-compatible backend…" replaces "Use this concept when documenting…"). No separate edit needed; the merge IS the fix.

…ickstart-polish Resolve PR #344's lingering mergeable: dirty state by linking the branch history to origin/main (2c1d485 = post-#326 / #348 / #330 / #341 / #343 / #345 / #346). The squash-merge commit landed earlier (e2af197) integrated main's content into the branch tree but did not link the histories, so GitHub's mergeable computation still saw the 3-way-merge-base artifact conflict on docs/quickstart/self-hosted.md (a vs A + the glossary link / "In another terminal" vs "Keep the gateway running" framing). This explicit merge commit ties the branch to main's history. Self-hosted.md conflict resolved by taking OUR side — the branch's edits already contain main's substantive changes (capital A, first-time-build paragraph) plus this PR's additions (glossary link, keep-running framing, YOUR_ADMIN_KEY note, config.yaml location anchor). The auto-merge of first-model-first-key-first-request.md duplicated the :::warning callout that was already integrated via the squash commit; removed the duplicate.

moonming · 2026-05-18T17:56:54Z

Summary

ai-gateway #345 (P2-A) added 11 long-tail OpenAI-adapter Provider::* variants + Hub registrations. The dashboard-side schema in schemas/resources/model.schema.json regenerated correctly. But there's a SEPARATELY-MAINTAINED hardcoded JSON schema in crates/aisix-core/src/models/schema.rs::model_schema() that backs aisix-etcd::loader's runtime validate_model check — and nobody updated it.

Result: cp-api admits a groq/mistral/etc. ProviderKey and Model (per #336 long-tail admission). The DP's aisix-etcd::loader reads the Model from etcd and validates against the OLD 6-value schema:

schema validation failed at `/provider`: "groq" is not one of ["openai","anthropic","google","deepseek","cohere","jina"]

Model rejected. Never appears in /v1/models. Customer's chat 404s on the alias.

Surfaced by AISIX-Cloud PR #366 e2e — first iteration of the D3.2 long-tail matrix. The groq Model creation succeeded at cp-api but the DP never saw it. Logs showed the schema rejection.

Fix

crates/aisix-core/src/models/schema.rs:120 — extend the provider enum from 6 to 17 values to match every Provider::as_str() output post-#345.

The existing negative-test model_unknown_provider_value_fails used "mistral" as the "unknown value" sentinel. Since mistral is now valid, renamed the fixture string to "this-is-not-a-provider-id" so the negative-test still pins the rejection class.

Why a separate schema even exists

aisix-etcd::loader::validate_and_parse runs schema validation BEFORE serde deserialization to fail-loud-and-skip on malformed entries (rather than crashing the DP). The schema is the gate; serde is the parser. Both must agree on the enum.

The dashboard-facing JSON schemas in schemas/resources/*.schema.json are GENERATED from the Rust types via cargo run --bin dump-schema (drift CI check enforces). The runtime validator's schema in schema.rs is hand-written and not gated by the drift check — a structural gap that this PR addresses with a behavioral fix only; a future refactor could fold the two sources together.

Test plan

cargo test -p aisix-core --lib schema — 52 tests pass
clippy clean
fmt clean
CI green
Audit (CLAUDE.md §7)
After merge + docker rebuild, AISIX-Cloud chore(fmt): restore rustfmt on aisix-etcd loader test (post-#362) #366 D3.2 batch 1 e2e exercises the unblocked path

References (per CLAUDE.md §7)

ai-gateway#345 (merged) — Provider enum + Hub register
aisix-etcd loader.rs:262 — the schema-rejection log path that surfaced the gap on PR chore(fmt): restore rustfmt on aisix-etcd loader test (post-#362) #366 CI

Refs api7/AISIX-Cloud#366

Summary by CodeRabbit

Improvements
- Expanded model provider validation to accept an extended set of provider identifiers during schema validation.
Tests
- Updated test cases to reflect the expanded provider validation rules, ensuring invalid providers continue to be properly rejected.

…ovider variants (#345 follow-up) ai-gateway #345 added 11 long-tail OpenAI-adapter Provider variants (groq, mistral, togetherai, fireworks-ai, perplexity, moonshotai, alibaba, zhipuai, baseten, huggingface, cerebras) to the Rust enum + Hub registrations + cp-api admission. The hardcoded JSON schema backing `aisix-etcd::loader`'s `validate_model` runtime check at `crates/aisix-core/src/models/schema.rs:120`, however, retained the original 6-value allowlist. The dashboard-side schema in `schemas/resources/model.schema.json` was regenerated correctly by #345's `dump-schema` step — that file IS now in sync with the enum. But the runtime validator uses a SEPARATELY-MAINTAINED hardcoded schema at line 112-200 of schema.rs that nobody updated. cp-api admission ✓, DP serde ✓, runtime schema validator ✗ — a customer's groq Model resource gets rejected at the etcd loader before deserialization with `schema validation failed at \`/provider\`: "groq" is not one of ["openai","anthropic","google","deepseek","cohere","jina"]`. This silently dropped EVERY long-tail Provider Model resource the dashboard wrote. The model never appears in `/v1/models` on the DP, and customer chat calls 404 on the alias. Surfaced by AISIX-Cloud PR #366 e2e (D3.2 batch 1) — the groq Model creation succeeds at cp-api but the DP never sees it. Logs showed the schema validation rejection. Fix `crates/aisix-core/src/models/schema.rs:120` — extend the `provider` enum to the full 17 values matching `Provider::as_str()` output across every variant: ```rust "provider": { "type": "string", "enum": [ "openai","anthropic","google","deepseek","cohere","jina", "groq","mistral","togetherai","fireworks-ai","perplexity", "moonshotai","alibaba","zhipuai","baseten","huggingface","cerebras" ] }, ``` The existing negative-test `model_unknown_provider_value_fails` used `"mistral"` as the "unknown value" sentinel; since mistral is now valid, the test would have stopped catching the rejection path. Renamed the test fixture value to `"this-is-not-a-provider-id"` so the negative-test still pins the rejection class. All 52 aisix-core schema tests pass; clippy clean; fmt clean. References (per CLAUDE.md §7) - ai-gateway#345 (merged) — Provider enum + Hub register additions - aisix-etcd loader.rs:262 — schema-rejection log path that surfaced the gap

coderabbitai · 2026-05-18T17:57:10Z

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Free

Run ID: 6608819e-114c-4760-8c03-7fa8bfd7b4c6

📥 Commits

Reviewing files that changed from the base of the PR and between 8d88b41 and ab76aa5.

📒 Files selected for processing (3)

crates/aisix-admin/src/lib.rs
crates/aisix-etcd/src/loader.rs
crates/aisix-etcd/src/supervisor.rs

✅ Files skipped from review due to trivial changes (1)

crates/aisix-etcd/src/supervisor.rs

📝 Walkthrough

Walkthrough

The PR expands the model JSON schema's provider field enum to accept a larger set of provider identifiers and updates tests/fixtures that previously used mistral to use a clearly invalid provider string so schema-rejection tests still fail.

Changes

Provider Field Allowlist Expansion

Layer / File(s)	Summary
Provider field schema and core validation test `crates/aisix-core/src/models/schema.rs`	The `provider` field enum in the model schema is expanded to accept a larger allowlist. The `model_unknown_provider_value_fails` test is updated to use an explicitly invalid provider string so the validation rejection path is still exercised.
Update tests and fixtures across crates `crates/aisix-admin/src/lib.rs`, `crates/aisix-etcd/src/loader.rs`, `crates/aisix-etcd/src/supervisor.rs`	Tests and JSON fixtures that previously used `provider: "mistral"` were changed to use `provider: "this-is-not-a-provider-id"` (or similar) to keep negative schema-validation tests failing under the expanded provider allowlist.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Note

🎁 Summarized by CodeRabbit Free

Your organization has reached its limit of developer seats under the Pro Plan. For new users, CodeRabbit will generate a high-level summary and a walkthrough for each pull request. For a comprehensive line-by-line review, please add seats to your subscription by visiting https://app.coderabbit.ai/login.If you believe this is a mistake and have available seats, please assign one to the pull request author through the subscription management page using the link above.

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

…rovider sentinel The negative-test fixtures across `aisix-admin` and `aisix-etcd` used `"mistral"` as the "unknown provider" sentinel — a vestige of the pre-#345 6-value allowlist. With ai-gateway#345 first-classing 11 long-tail variants (including mistral), those tests started asserting against a now-VALID provider and FAILED. Replaced every fixture occurrence with `"this-is-not-a-provider-id"`, which is intentionally NOT in the post-#345 17-value enum, keeping the schema-rejection class pinned without churn-tracking the live provider list. Files touched (post-#345 surgical sentinel rotation): - `crates/aisix-admin/src/lib.rs::create_model_with_invalid_provider_prefix_is_400_schema_error` - `crates/aisix-etcd/src/loader.rs::tests` (2 sites) - `crates/aisix-etcd/src/supervisor.rs::tests` (1 site, drives ~6 transitive supervisor test failures via shared fixture) All workspace tests green; clippy clean; fmt clean. Refs ai-gateway#345 ai-gateway#346

Integrate origin/main (commit 2c1d485 = post-PR-#326 / #348 plus #330 / #341 / #343 / #345 / #346) into this branch via `git merge --squash` to clear PR #344's lingering `mergeable: dirty` state. Conflict on `docs/quickstart/self-hosted.md` was a 3-way-merge-base artifact: base (3596c0a) read `- a reachable etcd instance`, main changed `a` → `A` (via #326), this branch additionally inserted the glossary link. Both changes are wanted; resolution per Umar's approved plan was `git checkout --ours`, which preserves the branch's self-hosted.md state (already integrates capital A + glossary link + first-time-build paragraph + keep-running framing). Other 4 overlapping doc files auto-merged cleanly (`bootstrap-config.md`, `core-concepts.md`, `first-model-first-key-first-request.md`, `openai-sdk.md`). Code files all auto-merged cleanly. Additional Copilot review (post-`167196a` cycle) addressed: - `docs/index.md:7` — change link display text from `[data-plane]` to `[data plane]` to match the canonical glossary term. The URL anchor `#data-plane` stays kebab-case (matches the glossary heading's auto-anchor); only the display text changes. Comment id 3271145422. - `docs/quickstart/openai-sdk.md:43` — change `All three steps below` to `All commands below`. The Install-the-SDK section has two command blocks (mkdir+cd, npm install), not three; the prior wording originated from a mental model (mkdir, cd, install) that doesn't match the typographic count of code blocks under the heading. Comment id 3271145458. Copilot's third comment on `docs/overview/core-concepts.md` Observability Exporter wording (id 3271145444) auto-resolves via this merge — main's #326 rewrite supersedes the branch's pre-#326 wording at that location ("ships per-request span telemetry… OTLP/HTTP-compatible backend…" replaces "Use this concept when documenting…"). No separate edit needed; the merge IS the fix.

…ickstart-polish Resolve PR #344's lingering mergeable: dirty state by linking the branch history to origin/main (2c1d485 = post-#326 / #348 / #330 / #341 / #343 / #345 / #346). The squash-merge commit landed earlier (e2af197) integrated main's content into the branch tree but did not link the histories, so GitHub's mergeable computation still saw the 3-way-merge-base artifact conflict on docs/quickstart/self-hosted.md (a vs A + the glossary link / "In another terminal" vs "Keep the gateway running" framing). This explicit merge commit ties the branch to main's history. Self-hosted.md conflict resolved by taking OUR side — the branch's edits already contain main's substantive changes (capital A, first-time-build paragraph) plus this PR's additions (glossary link, keep-running framing, YOUR_ADMIN_KEY note, config.yaml location anchor). The auto-merge of first-model-first-key-first-request.md duplicated the :::warning callout that was already integrated via the squash commit; removed the duplicate.

moonming · 2026-05-18T17:56:54Z

Summary

ai-gateway #345 (P2-A) added 11 long-tail OpenAI-adapter Provider::* variants + Hub registrations. The dashboard-side schema in schemas/resources/model.schema.json regenerated correctly. But there's a SEPARATELY-MAINTAINED hardcoded JSON schema in crates/aisix-core/src/models/schema.rs::model_schema() that backs aisix-etcd::loader's runtime validate_model check — and nobody updated it.

Result: cp-api admits a groq/mistral/etc. ProviderKey and Model (per #336 long-tail admission). The DP's aisix-etcd::loader reads the Model from etcd and validates against the OLD 6-value schema:

schema validation failed at `/provider`: "groq" is not one of ["openai","anthropic","google","deepseek","cohere","jina"]

Model rejected. Never appears in /v1/models. Customer's chat 404s on the alias.

Surfaced by AISIX-Cloud PR #366 e2e — first iteration of the D3.2 long-tail matrix. The groq Model creation succeeded at cp-api but the DP never saw it. Logs showed the schema rejection.

Fix

crates/aisix-core/src/models/schema.rs:120 — extend the provider enum from 6 to 17 values to match every Provider::as_str() output post-#345.

The existing negative-test model_unknown_provider_value_fails used "mistral" as the "unknown value" sentinel. Since mistral is now valid, renamed the fixture string to "this-is-not-a-provider-id" so the negative-test still pins the rejection class.

Why a separate schema even exists

aisix-etcd::loader::validate_and_parse runs schema validation BEFORE serde deserialization to fail-loud-and-skip on malformed entries (rather than crashing the DP). The schema is the gate; serde is the parser. Both must agree on the enum.

The dashboard-facing JSON schemas in schemas/resources/*.schema.json are GENERATED from the Rust types via cargo run --bin dump-schema (drift CI check enforces). The runtime validator's schema in schema.rs is hand-written and not gated by the drift check — a structural gap that this PR addresses with a behavioral fix only; a future refactor could fold the two sources together.

Test plan

cargo test -p aisix-core --lib schema — 52 tests pass
clippy clean
fmt clean
CI green
Audit (CLAUDE.md §7)
After merge + docker rebuild, AISIX-Cloud chore(fmt): restore rustfmt on aisix-etcd loader test (post-#362) #366 D3.2 batch 1 e2e exercises the unblocked path

References (per CLAUDE.md §7)

ai-gateway#345 (merged) — Provider enum + Hub register
aisix-etcd loader.rs:262 — the schema-rejection log path that surfaced the gap on PR chore(fmt): restore rustfmt on aisix-etcd loader test (post-#362) #366 CI

Refs api7/AISIX-Cloud#366

Summary by CodeRabbit

Improvements
- Expanded model provider validation to accept an extended set of provider identifiers during schema validation.
Tests
- Updated test cases to reflect the expanded provider validation rules, ensuring invalid providers continue to be properly rejected.

…ovider variants (#345 follow-up) ai-gateway #345 added 11 long-tail OpenAI-adapter Provider variants (groq, mistral, togetherai, fireworks-ai, perplexity, moonshotai, alibaba, zhipuai, baseten, huggingface, cerebras) to the Rust enum + Hub registrations + cp-api admission. The hardcoded JSON schema backing `aisix-etcd::loader`'s `validate_model` runtime check at `crates/aisix-core/src/models/schema.rs:120`, however, retained the original 6-value allowlist. The dashboard-side schema in `schemas/resources/model.schema.json` was regenerated correctly by #345's `dump-schema` step — that file IS now in sync with the enum. But the runtime validator uses a SEPARATELY-MAINTAINED hardcoded schema at line 112-200 of schema.rs that nobody updated. cp-api admission ✓, DP serde ✓, runtime schema validator ✗ — a customer's groq Model resource gets rejected at the etcd loader before deserialization with `schema validation failed at \`/provider\`: "groq" is not one of ["openai","anthropic","google","deepseek","cohere","jina"]`. This silently dropped EVERY long-tail Provider Model resource the dashboard wrote. The model never appears in `/v1/models` on the DP, and customer chat calls 404 on the alias. Surfaced by AISIX-Cloud PR #366 e2e (D3.2 batch 1) — the groq Model creation succeeds at cp-api but the DP never sees it. Logs showed the schema validation rejection. Fix `crates/aisix-core/src/models/schema.rs:120` — extend the `provider` enum to the full 17 values matching `Provider::as_str()` output across every variant: ```rust "provider": { "type": "string", "enum": [ "openai","anthropic","google","deepseek","cohere","jina", "groq","mistral","togetherai","fireworks-ai","perplexity", "moonshotai","alibaba","zhipuai","baseten","huggingface","cerebras" ] }, ``` The existing negative-test `model_unknown_provider_value_fails` used `"mistral"` as the "unknown value" sentinel; since mistral is now valid, the test would have stopped catching the rejection path. Renamed the test fixture value to `"this-is-not-a-provider-id"` so the negative-test still pins the rejection class. All 52 aisix-core schema tests pass; clippy clean; fmt clean. References (per CLAUDE.md §7) - ai-gateway#345 (merged) — Provider enum + Hub register additions - aisix-etcd loader.rs:262 — schema-rejection log path that surfaced the gap

coderabbitai · 2026-05-18T17:57:10Z

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Free

Run ID: 6608819e-114c-4760-8c03-7fa8bfd7b4c6

📥 Commits

Reviewing files that changed from the base of the PR and between 8d88b41 and ab76aa5.

📒 Files selected for processing (3)

crates/aisix-admin/src/lib.rs
crates/aisix-etcd/src/loader.rs
crates/aisix-etcd/src/supervisor.rs

✅ Files skipped from review due to trivial changes (1)

crates/aisix-etcd/src/supervisor.rs

📝 Walkthrough

Walkthrough

The PR expands the model JSON schema's provider field enum to accept a larger set of provider identifiers and updates tests/fixtures that previously used mistral to use a clearly invalid provider string so schema-rejection tests still fail.

Changes

Provider Field Allowlist Expansion

Layer / File(s)	Summary
Provider field schema and core validation test `crates/aisix-core/src/models/schema.rs`	The `provider` field enum in the model schema is expanded to accept a larger allowlist. The `model_unknown_provider_value_fails` test is updated to use an explicitly invalid provider string so the validation rejection path is still exercised.
Update tests and fixtures across crates `crates/aisix-admin/src/lib.rs`, `crates/aisix-etcd/src/loader.rs`, `crates/aisix-etcd/src/supervisor.rs`	Tests and JSON fixtures that previously used `provider: "mistral"` were changed to use `provider: "this-is-not-a-provider-id"` (or similar) to keep negative schema-validation tests failing under the expanded provider allowlist.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Note

🎁 Summarized by CodeRabbit Free

Your organization has reached its limit of developer seats under the Pro Plan. For new users, CodeRabbit will generate a high-level summary and a walkthrough for each pull request. For a comprehensive line-by-line review, please add seats to your subscription by visiting https://app.coderabbit.ai/login.If you believe this is a mistake and have available seats, please assign one to the pull request author through the subscription management page using the link above.

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

…rovider sentinel The negative-test fixtures across `aisix-admin` and `aisix-etcd` used `"mistral"` as the "unknown provider" sentinel — a vestige of the pre-#345 6-value allowlist. With ai-gateway#345 first-classing 11 long-tail variants (including mistral), those tests started asserting against a now-VALID provider and FAILED. Replaced every fixture occurrence with `"this-is-not-a-provider-id"`, which is intentionally NOT in the post-#345 17-value enum, keeping the schema-rejection class pinned without churn-tracking the live provider list. Files touched (post-#345 surgical sentinel rotation): - `crates/aisix-admin/src/lib.rs::create_model_with_invalid_provider_prefix_is_400_schema_error` - `crates/aisix-etcd/src/loader.rs::tests` (2 sites) - `crates/aisix-etcd/src/supervisor.rs::tests` (1 site, drives ~6 transitive supervisor test failures via shared fixture) All workspace tests green; clippy clean; fmt clean. Refs ai-gateway#345 ai-gateway#346

Integrate origin/main (commit 2c1d485 = post-PR-#326 / #348 plus #330 / #341 / #343 / #345 / #346) into this branch via `git merge --squash` to clear PR #344's lingering `mergeable: dirty` state. Conflict on `docs/quickstart/self-hosted.md` was a 3-way-merge-base artifact: base (3596c0a) read `- a reachable etcd instance`, main changed `a` → `A` (via #326), this branch additionally inserted the glossary link. Both changes are wanted; resolution per Umar's approved plan was `git checkout --ours`, which preserves the branch's self-hosted.md state (already integrates capital A + glossary link + first-time-build paragraph + keep-running framing). Other 4 overlapping doc files auto-merged cleanly (`bootstrap-config.md`, `core-concepts.md`, `first-model-first-key-first-request.md`, `openai-sdk.md`). Code files all auto-merged cleanly. Additional Copilot review (post-`167196a` cycle) addressed: - `docs/index.md:7` — change link display text from `[data-plane]` to `[data plane]` to match the canonical glossary term. The URL anchor `#data-plane` stays kebab-case (matches the glossary heading's auto-anchor); only the display text changes. Comment id 3271145422. - `docs/quickstart/openai-sdk.md:43` — change `All three steps below` to `All commands below`. The Install-the-SDK section has two command blocks (mkdir+cd, npm install), not three; the prior wording originated from a mental model (mkdir, cd, install) that doesn't match the typographic count of code blocks under the heading. Comment id 3271145458. Copilot's third comment on `docs/overview/core-concepts.md` Observability Exporter wording (id 3271145444) auto-resolves via this merge — main's #326 rewrite supersedes the branch's pre-#326 wording at that location ("ships per-request span telemetry… OTLP/HTTP-compatible backend…" replaces "Use this concept when documenting…"). No separate edit needed; the merge IS the fix.

…ickstart-polish Resolve PR #344's lingering mergeable: dirty state by linking the branch history to origin/main (2c1d485 = post-#326 / #348 / #330 / #341 / #343 / #345 / #346). The squash-merge commit landed earlier (e2af197) integrated main's content into the branch tree but did not link the histories, so GitHub's mergeable computation still saw the 3-way-merge-base artifact conflict on docs/quickstart/self-hosted.md (a vs A + the glossary link / "In another terminal" vs "Keep the gateway running" framing). This explicit merge commit ties the branch to main's history. Self-hosted.md conflict resolved by taking OUR side — the branch's edits already contain main's substantive changes (capital A, first-time-build paragraph) plus this PR's additions (glossary link, keep-running framing, YOUR_ADMIN_KEY note, config.yaml location anchor). The auto-merge of first-model-first-key-first-request.md duplicated the :::warning callout that was already integrated via the squash commit; removed the duplicate.

moonming · 2026-05-18T17:56:54Z

Summary

ai-gateway #345 (P2-A) added 11 long-tail OpenAI-adapter Provider::* variants + Hub registrations. The dashboard-side schema in schemas/resources/model.schema.json regenerated correctly. But there's a SEPARATELY-MAINTAINED hardcoded JSON schema in crates/aisix-core/src/models/schema.rs::model_schema() that backs aisix-etcd::loader's runtime validate_model check — and nobody updated it.

Result: cp-api admits a groq/mistral/etc. ProviderKey and Model (per #336 long-tail admission). The DP's aisix-etcd::loader reads the Model from etcd and validates against the OLD 6-value schema:

schema validation failed at `/provider`: "groq" is not one of ["openai","anthropic","google","deepseek","cohere","jina"]

Model rejected. Never appears in /v1/models. Customer's chat 404s on the alias.

Surfaced by AISIX-Cloud PR #366 e2e — first iteration of the D3.2 long-tail matrix. The groq Model creation succeeded at cp-api but the DP never saw it. Logs showed the schema rejection.

Fix

crates/aisix-core/src/models/schema.rs:120 — extend the provider enum from 6 to 17 values to match every Provider::as_str() output post-#345.

The existing negative-test model_unknown_provider_value_fails used "mistral" as the "unknown value" sentinel. Since mistral is now valid, renamed the fixture string to "this-is-not-a-provider-id" so the negative-test still pins the rejection class.

Why a separate schema even exists

aisix-etcd::loader::validate_and_parse runs schema validation BEFORE serde deserialization to fail-loud-and-skip on malformed entries (rather than crashing the DP). The schema is the gate; serde is the parser. Both must agree on the enum.

The dashboard-facing JSON schemas in schemas/resources/*.schema.json are GENERATED from the Rust types via cargo run --bin dump-schema (drift CI check enforces). The runtime validator's schema in schema.rs is hand-written and not gated by the drift check — a structural gap that this PR addresses with a behavioral fix only; a future refactor could fold the two sources together.

Test plan

cargo test -p aisix-core --lib schema — 52 tests pass
clippy clean
fmt clean
CI green
Audit (CLAUDE.md §7)
After merge + docker rebuild, AISIX-Cloud chore(fmt): restore rustfmt on aisix-etcd loader test (post-#362) #366 D3.2 batch 1 e2e exercises the unblocked path

References (per CLAUDE.md §7)

ai-gateway#345 (merged) — Provider enum + Hub register
aisix-etcd loader.rs:262 — the schema-rejection log path that surfaced the gap on PR chore(fmt): restore rustfmt on aisix-etcd loader test (post-#362) #366 CI

Refs api7/AISIX-Cloud#366

Summary by CodeRabbit

Improvements
- Expanded model provider validation to accept an extended set of provider identifiers during schema validation.
Tests
- Updated test cases to reflect the expanded provider validation rules, ensuring invalid providers continue to be properly rejected.

…ovider variants (#345 follow-up) ai-gateway #345 added 11 long-tail OpenAI-adapter Provider variants (groq, mistral, togetherai, fireworks-ai, perplexity, moonshotai, alibaba, zhipuai, baseten, huggingface, cerebras) to the Rust enum + Hub registrations + cp-api admission. The hardcoded JSON schema backing `aisix-etcd::loader`'s `validate_model` runtime check at `crates/aisix-core/src/models/schema.rs:120`, however, retained the original 6-value allowlist. The dashboard-side schema in `schemas/resources/model.schema.json` was regenerated correctly by #345's `dump-schema` step — that file IS now in sync with the enum. But the runtime validator uses a SEPARATELY-MAINTAINED hardcoded schema at line 112-200 of schema.rs that nobody updated. cp-api admission ✓, DP serde ✓, runtime schema validator ✗ — a customer's groq Model resource gets rejected at the etcd loader before deserialization with `schema validation failed at \`/provider\`: "groq" is not one of ["openai","anthropic","google","deepseek","cohere","jina"]`. This silently dropped EVERY long-tail Provider Model resource the dashboard wrote. The model never appears in `/v1/models` on the DP, and customer chat calls 404 on the alias. Surfaced by AISIX-Cloud PR #366 e2e (D3.2 batch 1) — the groq Model creation succeeds at cp-api but the DP never sees it. Logs showed the schema validation rejection. Fix `crates/aisix-core/src/models/schema.rs:120` — extend the `provider` enum to the full 17 values matching `Provider::as_str()` output across every variant: ```rust "provider": { "type": "string", "enum": [ "openai","anthropic","google","deepseek","cohere","jina", "groq","mistral","togetherai","fireworks-ai","perplexity", "moonshotai","alibaba","zhipuai","baseten","huggingface","cerebras" ] }, ``` The existing negative-test `model_unknown_provider_value_fails` used `"mistral"` as the "unknown value" sentinel; since mistral is now valid, the test would have stopped catching the rejection path. Renamed the test fixture value to `"this-is-not-a-provider-id"` so the negative-test still pins the rejection class. All 52 aisix-core schema tests pass; clippy clean; fmt clean. References (per CLAUDE.md §7) - ai-gateway#345 (merged) — Provider enum + Hub register additions - aisix-etcd loader.rs:262 — schema-rejection log path that surfaced the gap

coderabbitai · 2026-05-18T17:57:10Z

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Free

Run ID: 6608819e-114c-4760-8c03-7fa8bfd7b4c6

📥 Commits

Reviewing files that changed from the base of the PR and between 8d88b41 and ab76aa5.

📒 Files selected for processing (3)

crates/aisix-admin/src/lib.rs
crates/aisix-etcd/src/loader.rs
crates/aisix-etcd/src/supervisor.rs

✅ Files skipped from review due to trivial changes (1)

crates/aisix-etcd/src/supervisor.rs

📝 Walkthrough

Walkthrough

The PR expands the model JSON schema's provider field enum to accept a larger set of provider identifiers and updates tests/fixtures that previously used mistral to use a clearly invalid provider string so schema-rejection tests still fail.

Changes

Provider Field Allowlist Expansion

Layer / File(s)	Summary
Provider field schema and core validation test `crates/aisix-core/src/models/schema.rs`	The `provider` field enum in the model schema is expanded to accept a larger allowlist. The `model_unknown_provider_value_fails` test is updated to use an explicitly invalid provider string so the validation rejection path is still exercised.
Update tests and fixtures across crates `crates/aisix-admin/src/lib.rs`, `crates/aisix-etcd/src/loader.rs`, `crates/aisix-etcd/src/supervisor.rs`	Tests and JSON fixtures that previously used `provider: "mistral"` were changed to use `provider: "this-is-not-a-provider-id"` (or similar) to keep negative schema-validation tests failing under the expanded provider allowlist.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Note

🎁 Summarized by CodeRabbit Free

Your organization has reached its limit of developer seats under the Pro Plan. For new users, CodeRabbit will generate a high-level summary and a walkthrough for each pull request. For a comprehensive line-by-line review, please add seats to your subscription by visiting https://app.coderabbit.ai/login.If you believe this is a mistake and have available seats, please assign one to the pull request author through the subscription management page using the link above.

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

…rovider sentinel The negative-test fixtures across `aisix-admin` and `aisix-etcd` used `"mistral"` as the "unknown provider" sentinel — a vestige of the pre-#345 6-value allowlist. With ai-gateway#345 first-classing 11 long-tail variants (including mistral), those tests started asserting against a now-VALID provider and FAILED. Replaced every fixture occurrence with `"this-is-not-a-provider-id"`, which is intentionally NOT in the post-#345 17-value enum, keeping the schema-rejection class pinned without churn-tracking the live provider list. Files touched (post-#345 surgical sentinel rotation): - `crates/aisix-admin/src/lib.rs::create_model_with_invalid_provider_prefix_is_400_schema_error` - `crates/aisix-etcd/src/loader.rs::tests` (2 sites) - `crates/aisix-etcd/src/supervisor.rs::tests` (1 site, drives ~6 transitive supervisor test failures via shared fixture) All workspace tests green; clippy clean; fmt clean. Refs ai-gateway#345 ai-gateway#346

Integrate origin/main (commit 2c1d485 = post-PR-#326 / #348 plus #330 / #341 / #343 / #345 / #346) into this branch via `git merge --squash` to clear PR #344's lingering `mergeable: dirty` state. Conflict on `docs/quickstart/self-hosted.md` was a 3-way-merge-base artifact: base (3596c0a) read `- a reachable etcd instance`, main changed `a` → `A` (via #326), this branch additionally inserted the glossary link. Both changes are wanted; resolution per Umar's approved plan was `git checkout --ours`, which preserves the branch's self-hosted.md state (already integrates capital A + glossary link + first-time-build paragraph + keep-running framing). Other 4 overlapping doc files auto-merged cleanly (`bootstrap-config.md`, `core-concepts.md`, `first-model-first-key-first-request.md`, `openai-sdk.md`). Code files all auto-merged cleanly. Additional Copilot review (post-`167196a` cycle) addressed: - `docs/index.md:7` — change link display text from `[data-plane]` to `[data plane]` to match the canonical glossary term. The URL anchor `#data-plane` stays kebab-case (matches the glossary heading's auto-anchor); only the display text changes. Comment id 3271145422. - `docs/quickstart/openai-sdk.md:43` — change `All three steps below` to `All commands below`. The Install-the-SDK section has two command blocks (mkdir+cd, npm install), not three; the prior wording originated from a mental model (mkdir, cd, install) that doesn't match the typographic count of code blocks under the heading. Comment id 3271145458. Copilot's third comment on `docs/overview/core-concepts.md` Observability Exporter wording (id 3271145444) auto-resolves via this merge — main's #326 rewrite supersedes the branch's pre-#326 wording at that location ("ships per-request span telemetry… OTLP/HTTP-compatible backend…" replaces "Use this concept when documenting…"). No separate edit needed; the merge IS the fix.

…ickstart-polish Resolve PR #344's lingering mergeable: dirty state by linking the branch history to origin/main (2c1d485 = post-#326 / #348 / #330 / #341 / #343 / #345 / #346). The squash-merge commit landed earlier (e2af197) integrated main's content into the branch tree but did not link the histories, so GitHub's mergeable computation still saw the 3-way-merge-base artifact conflict on docs/quickstart/self-hosted.md (a vs A + the glossary link / "In another terminal" vs "Keep the gateway running" framing). This explicit merge commit ties the branch to main's history. Self-hosted.md conflict resolved by taking OUR side — the branch's edits already contain main's substantive changes (capital A, first-time-build paragraph) plus this PR's additions (glossary link, keep-running framing, YOUR_ADMIN_KEY note, config.yaml location anchor). The auto-merge of first-model-first-key-first-request.md duplicated the :::warning callout that was already integrated via the squash commit; removed the duplicate.

moonming · 2026-05-18T17:56:54Z

Summary

ai-gateway #345 (P2-A) added 11 long-tail OpenAI-adapter Provider::* variants + Hub registrations. The dashboard-side schema in schemas/resources/model.schema.json regenerated correctly. But there's a SEPARATELY-MAINTAINED hardcoded JSON schema in crates/aisix-core/src/models/schema.rs::model_schema() that backs aisix-etcd::loader's runtime validate_model check — and nobody updated it.

Result: cp-api admits a groq/mistral/etc. ProviderKey and Model (per #336 long-tail admission). The DP's aisix-etcd::loader reads the Model from etcd and validates against the OLD 6-value schema:

schema validation failed at `/provider`: "groq" is not one of ["openai","anthropic","google","deepseek","cohere","jina"]

Model rejected. Never appears in /v1/models. Customer's chat 404s on the alias.

Surfaced by AISIX-Cloud PR #366 e2e — first iteration of the D3.2 long-tail matrix. The groq Model creation succeeded at cp-api but the DP never saw it. Logs showed the schema rejection.

Fix

crates/aisix-core/src/models/schema.rs:120 — extend the provider enum from 6 to 17 values to match every Provider::as_str() output post-#345.

The existing negative-test model_unknown_provider_value_fails used "mistral" as the "unknown value" sentinel. Since mistral is now valid, renamed the fixture string to "this-is-not-a-provider-id" so the negative-test still pins the rejection class.

Why a separate schema even exists

aisix-etcd::loader::validate_and_parse runs schema validation BEFORE serde deserialization to fail-loud-and-skip on malformed entries (rather than crashing the DP). The schema is the gate; serde is the parser. Both must agree on the enum.

The dashboard-facing JSON schemas in schemas/resources/*.schema.json are GENERATED from the Rust types via cargo run --bin dump-schema (drift CI check enforces). The runtime validator's schema in schema.rs is hand-written and not gated by the drift check — a structural gap that this PR addresses with a behavioral fix only; a future refactor could fold the two sources together.

Test plan

cargo test -p aisix-core --lib schema — 52 tests pass
clippy clean
fmt clean
CI green
Audit (CLAUDE.md §7)
After merge + docker rebuild, AISIX-Cloud chore(fmt): restore rustfmt on aisix-etcd loader test (post-#362) #366 D3.2 batch 1 e2e exercises the unblocked path

References (per CLAUDE.md §7)

ai-gateway#345 (merged) — Provider enum + Hub register
aisix-etcd loader.rs:262 — the schema-rejection log path that surfaced the gap on PR chore(fmt): restore rustfmt on aisix-etcd loader test (post-#362) #366 CI

Refs api7/AISIX-Cloud#366

Summary by CodeRabbit

Improvements
- Expanded model provider validation to accept an extended set of provider identifiers during schema validation.
Tests
- Updated test cases to reflect the expanded provider validation rules, ensuring invalid providers continue to be properly rejected.

…ovider variants (#345 follow-up) ai-gateway #345 added 11 long-tail OpenAI-adapter Provider variants (groq, mistral, togetherai, fireworks-ai, perplexity, moonshotai, alibaba, zhipuai, baseten, huggingface, cerebras) to the Rust enum + Hub registrations + cp-api admission. The hardcoded JSON schema backing `aisix-etcd::loader`'s `validate_model` runtime check at `crates/aisix-core/src/models/schema.rs:120`, however, retained the original 6-value allowlist. The dashboard-side schema in `schemas/resources/model.schema.json` was regenerated correctly by #345's `dump-schema` step — that file IS now in sync with the enum. But the runtime validator uses a SEPARATELY-MAINTAINED hardcoded schema at line 112-200 of schema.rs that nobody updated. cp-api admission ✓, DP serde ✓, runtime schema validator ✗ — a customer's groq Model resource gets rejected at the etcd loader before deserialization with `schema validation failed at \`/provider\`: "groq" is not one of ["openai","anthropic","google","deepseek","cohere","jina"]`. This silently dropped EVERY long-tail Provider Model resource the dashboard wrote. The model never appears in `/v1/models` on the DP, and customer chat calls 404 on the alias. Surfaced by AISIX-Cloud PR #366 e2e (D3.2 batch 1) — the groq Model creation succeeds at cp-api but the DP never sees it. Logs showed the schema validation rejection. Fix `crates/aisix-core/src/models/schema.rs:120` — extend the `provider` enum to the full 17 values matching `Provider::as_str()` output across every variant: ```rust "provider": { "type": "string", "enum": [ "openai","anthropic","google","deepseek","cohere","jina", "groq","mistral","togetherai","fireworks-ai","perplexity", "moonshotai","alibaba","zhipuai","baseten","huggingface","cerebras" ] }, ``` The existing negative-test `model_unknown_provider_value_fails` used `"mistral"` as the "unknown value" sentinel; since mistral is now valid, the test would have stopped catching the rejection path. Renamed the test fixture value to `"this-is-not-a-provider-id"` so the negative-test still pins the rejection class. All 52 aisix-core schema tests pass; clippy clean; fmt clean. References (per CLAUDE.md §7) - ai-gateway#345 (merged) — Provider enum + Hub register additions - aisix-etcd loader.rs:262 — schema-rejection log path that surfaced the gap

coderabbitai · 2026-05-18T17:57:10Z

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Free

Run ID: 6608819e-114c-4760-8c03-7fa8bfd7b4c6

📥 Commits

Reviewing files that changed from the base of the PR and between 8d88b41 and ab76aa5.

📒 Files selected for processing (3)

crates/aisix-admin/src/lib.rs
crates/aisix-etcd/src/loader.rs
crates/aisix-etcd/src/supervisor.rs

✅ Files skipped from review due to trivial changes (1)

crates/aisix-etcd/src/supervisor.rs

📝 Walkthrough

Walkthrough

The PR expands the model JSON schema's provider field enum to accept a larger set of provider identifiers and updates tests/fixtures that previously used mistral to use a clearly invalid provider string so schema-rejection tests still fail.

Changes

Provider Field Allowlist Expansion

Layer / File(s)	Summary
Provider field schema and core validation test `crates/aisix-core/src/models/schema.rs`	The `provider` field enum in the model schema is expanded to accept a larger allowlist. The `model_unknown_provider_value_fails` test is updated to use an explicitly invalid provider string so the validation rejection path is still exercised.
Update tests and fixtures across crates `crates/aisix-admin/src/lib.rs`, `crates/aisix-etcd/src/loader.rs`, `crates/aisix-etcd/src/supervisor.rs`	Tests and JSON fixtures that previously used `provider: "mistral"` were changed to use `provider: "this-is-not-a-provider-id"` (or similar) to keep negative schema-validation tests failing under the expanded provider allowlist.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Note

🎁 Summarized by CodeRabbit Free

Your organization has reached its limit of developer seats under the Pro Plan. For new users, CodeRabbit will generate a high-level summary and a walkthrough for each pull request. For a comprehensive line-by-line review, please add seats to your subscription by visiting https://app.coderabbit.ai/login.If you believe this is a mistake and have available seats, please assign one to the pull request author through the subscription management page using the link above.

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

…rovider sentinel The negative-test fixtures across `aisix-admin` and `aisix-etcd` used `"mistral"` as the "unknown provider" sentinel — a vestige of the pre-#345 6-value allowlist. With ai-gateway#345 first-classing 11 long-tail variants (including mistral), those tests started asserting against a now-VALID provider and FAILED. Replaced every fixture occurrence with `"this-is-not-a-provider-id"`, which is intentionally NOT in the post-#345 17-value enum, keeping the schema-rejection class pinned without churn-tracking the live provider list. Files touched (post-#345 surgical sentinel rotation): - `crates/aisix-admin/src/lib.rs::create_model_with_invalid_provider_prefix_is_400_schema_error` - `crates/aisix-etcd/src/loader.rs::tests` (2 sites) - `crates/aisix-etcd/src/supervisor.rs::tests` (1 site, drives ~6 transitive supervisor test failures via shared fixture) All workspace tests green; clippy clean; fmt clean. Refs ai-gateway#345 ai-gateway#346

Integrate origin/main (commit 2c1d485 = post-PR-#326 / #348 plus #330 / #341 / #343 / #345 / #346) into this branch via `git merge --squash` to clear PR #344's lingering `mergeable: dirty` state. Conflict on `docs/quickstart/self-hosted.md` was a 3-way-merge-base artifact: base (3596c0a) read `- a reachable etcd instance`, main changed `a` → `A` (via #326), this branch additionally inserted the glossary link. Both changes are wanted; resolution per Umar's approved plan was `git checkout --ours`, which preserves the branch's self-hosted.md state (already integrates capital A + glossary link + first-time-build paragraph + keep-running framing). Other 4 overlapping doc files auto-merged cleanly (`bootstrap-config.md`, `core-concepts.md`, `first-model-first-key-first-request.md`, `openai-sdk.md`). Code files all auto-merged cleanly. Additional Copilot review (post-`167196a` cycle) addressed: - `docs/index.md:7` — change link display text from `[data-plane]` to `[data plane]` to match the canonical glossary term. The URL anchor `#data-plane` stays kebab-case (matches the glossary heading's auto-anchor); only the display text changes. Comment id 3271145422. - `docs/quickstart/openai-sdk.md:43` — change `All three steps below` to `All commands below`. The Install-the-SDK section has two command blocks (mkdir+cd, npm install), not three; the prior wording originated from a mental model (mkdir, cd, install) that doesn't match the typographic count of code blocks under the heading. Comment id 3271145458. Copilot's third comment on `docs/overview/core-concepts.md` Observability Exporter wording (id 3271145444) auto-resolves via this merge — main's #326 rewrite supersedes the branch's pre-#326 wording at that location ("ships per-request span telemetry… OTLP/HTTP-compatible backend…" replaces "Use this concept when documenting…"). No separate edit needed; the merge IS the fix.

…ickstart-polish Resolve PR #344's lingering mergeable: dirty state by linking the branch history to origin/main (2c1d485 = post-#326 / #348 / #330 / #341 / #343 / #345 / #346). The squash-merge commit landed earlier (e2af197) integrated main's content into the branch tree but did not link the histories, so GitHub's mergeable computation still saw the 3-way-merge-base artifact conflict on docs/quickstart/self-hosted.md (a vs A + the glossary link / "In another terminal" vs "Keep the gateway running" framing). This explicit merge commit ties the branch to main's history. Self-hosted.md conflict resolved by taking OUR side — the branch's edits already contain main's substantive changes (capital A, first-time-build paragraph) plus this PR's additions (glossary link, keep-running framing, YOUR_ADMIN_KEY note, config.yaml location anchor). The auto-merge of first-model-first-key-first-request.md duplicated the :::warning callout that was already integrated via the squash commit; removed the duplicate.

moonming · 2026-05-18T17:56:54Z

Summary

ai-gateway #345 (P2-A) added 11 long-tail OpenAI-adapter Provider::* variants + Hub registrations. The dashboard-side schema in schemas/resources/model.schema.json regenerated correctly. But there's a SEPARATELY-MAINTAINED hardcoded JSON schema in crates/aisix-core/src/models/schema.rs::model_schema() that backs aisix-etcd::loader's runtime validate_model check — and nobody updated it.

Result: cp-api admits a groq/mistral/etc. ProviderKey and Model (per #336 long-tail admission). The DP's aisix-etcd::loader reads the Model from etcd and validates against the OLD 6-value schema:

schema validation failed at `/provider`: "groq" is not one of ["openai","anthropic","google","deepseek","cohere","jina"]

Model rejected. Never appears in /v1/models. Customer's chat 404s on the alias.

Surfaced by AISIX-Cloud PR #366 e2e — first iteration of the D3.2 long-tail matrix. The groq Model creation succeeded at cp-api but the DP never saw it. Logs showed the schema rejection.

Fix

crates/aisix-core/src/models/schema.rs:120 — extend the provider enum from 6 to 17 values to match every Provider::as_str() output post-#345.

The existing negative-test model_unknown_provider_value_fails used "mistral" as the "unknown value" sentinel. Since mistral is now valid, renamed the fixture string to "this-is-not-a-provider-id" so the negative-test still pins the rejection class.

Why a separate schema even exists

aisix-etcd::loader::validate_and_parse runs schema validation BEFORE serde deserialization to fail-loud-and-skip on malformed entries (rather than crashing the DP). The schema is the gate; serde is the parser. Both must agree on the enum.

The dashboard-facing JSON schemas in schemas/resources/*.schema.json are GENERATED from the Rust types via cargo run --bin dump-schema (drift CI check enforces). The runtime validator's schema in schema.rs is hand-written and not gated by the drift check — a structural gap that this PR addresses with a behavioral fix only; a future refactor could fold the two sources together.

Test plan

cargo test -p aisix-core --lib schema — 52 tests pass
clippy clean
fmt clean
CI green
Audit (CLAUDE.md §7)
After merge + docker rebuild, AISIX-Cloud chore(fmt): restore rustfmt on aisix-etcd loader test (post-#362) #366 D3.2 batch 1 e2e exercises the unblocked path

References (per CLAUDE.md §7)

ai-gateway#345 (merged) — Provider enum + Hub register
aisix-etcd loader.rs:262 — the schema-rejection log path that surfaced the gap on PR chore(fmt): restore rustfmt on aisix-etcd loader test (post-#362) #366 CI

Refs api7/AISIX-Cloud#366

Summary by CodeRabbit

Improvements
- Expanded model provider validation to accept an extended set of provider identifiers during schema validation.
Tests
- Updated test cases to reflect the expanded provider validation rules, ensuring invalid providers continue to be properly rejected.

…ovider variants (#345 follow-up) ai-gateway #345 added 11 long-tail OpenAI-adapter Provider variants (groq, mistral, togetherai, fireworks-ai, perplexity, moonshotai, alibaba, zhipuai, baseten, huggingface, cerebras) to the Rust enum + Hub registrations + cp-api admission. The hardcoded JSON schema backing `aisix-etcd::loader`'s `validate_model` runtime check at `crates/aisix-core/src/models/schema.rs:120`, however, retained the original 6-value allowlist. The dashboard-side schema in `schemas/resources/model.schema.json` was regenerated correctly by #345's `dump-schema` step — that file IS now in sync with the enum. But the runtime validator uses a SEPARATELY-MAINTAINED hardcoded schema at line 112-200 of schema.rs that nobody updated. cp-api admission ✓, DP serde ✓, runtime schema validator ✗ — a customer's groq Model resource gets rejected at the etcd loader before deserialization with `schema validation failed at \`/provider\`: "groq" is not one of ["openai","anthropic","google","deepseek","cohere","jina"]`. This silently dropped EVERY long-tail Provider Model resource the dashboard wrote. The model never appears in `/v1/models` on the DP, and customer chat calls 404 on the alias. Surfaced by AISIX-Cloud PR #366 e2e (D3.2 batch 1) — the groq Model creation succeeds at cp-api but the DP never sees it. Logs showed the schema validation rejection. Fix `crates/aisix-core/src/models/schema.rs:120` — extend the `provider` enum to the full 17 values matching `Provider::as_str()` output across every variant: ```rust "provider": { "type": "string", "enum": [ "openai","anthropic","google","deepseek","cohere","jina", "groq","mistral","togetherai","fireworks-ai","perplexity", "moonshotai","alibaba","zhipuai","baseten","huggingface","cerebras" ] }, ``` The existing negative-test `model_unknown_provider_value_fails` used `"mistral"` as the "unknown value" sentinel; since mistral is now valid, the test would have stopped catching the rejection path. Renamed the test fixture value to `"this-is-not-a-provider-id"` so the negative-test still pins the rejection class. All 52 aisix-core schema tests pass; clippy clean; fmt clean. References (per CLAUDE.md §7) - ai-gateway#345 (merged) — Provider enum + Hub register additions - aisix-etcd loader.rs:262 — schema-rejection log path that surfaced the gap

coderabbitai · 2026-05-18T17:57:10Z

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Free

Run ID: 6608819e-114c-4760-8c03-7fa8bfd7b4c6

📥 Commits

Reviewing files that changed from the base of the PR and between 8d88b41 and ab76aa5.

📒 Files selected for processing (3)

crates/aisix-admin/src/lib.rs
crates/aisix-etcd/src/loader.rs
crates/aisix-etcd/src/supervisor.rs

✅ Files skipped from review due to trivial changes (1)

crates/aisix-etcd/src/supervisor.rs

📝 Walkthrough

Walkthrough

The PR expands the model JSON schema's provider field enum to accept a larger set of provider identifiers and updates tests/fixtures that previously used mistral to use a clearly invalid provider string so schema-rejection tests still fail.

Changes

Provider Field Allowlist Expansion

Layer / File(s)	Summary
Provider field schema and core validation test `crates/aisix-core/src/models/schema.rs`	The `provider` field enum in the model schema is expanded to accept a larger allowlist. The `model_unknown_provider_value_fails` test is updated to use an explicitly invalid provider string so the validation rejection path is still exercised.
Update tests and fixtures across crates `crates/aisix-admin/src/lib.rs`, `crates/aisix-etcd/src/loader.rs`, `crates/aisix-etcd/src/supervisor.rs`	Tests and JSON fixtures that previously used `provider: "mistral"` were changed to use `provider: "this-is-not-a-provider-id"` (or similar) to keep negative schema-validation tests failing under the expanded provider allowlist.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Note

🎁 Summarized by CodeRabbit Free

Your organization has reached its limit of developer seats under the Pro Plan. For new users, CodeRabbit will generate a high-level summary and a walkthrough for each pull request. For a comprehensive line-by-line review, please add seats to your subscription by visiting https://app.coderabbit.ai/login.If you believe this is a mistake and have available seats, please assign one to the pull request author through the subscription management page using the link above.

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

…rovider sentinel The negative-test fixtures across `aisix-admin` and `aisix-etcd` used `"mistral"` as the "unknown provider" sentinel — a vestige of the pre-#345 6-value allowlist. With ai-gateway#345 first-classing 11 long-tail variants (including mistral), those tests started asserting against a now-VALID provider and FAILED. Replaced every fixture occurrence with `"this-is-not-a-provider-id"`, which is intentionally NOT in the post-#345 17-value enum, keeping the schema-rejection class pinned without churn-tracking the live provider list. Files touched (post-#345 surgical sentinel rotation): - `crates/aisix-admin/src/lib.rs::create_model_with_invalid_provider_prefix_is_400_schema_error` - `crates/aisix-etcd/src/loader.rs::tests` (2 sites) - `crates/aisix-etcd/src/supervisor.rs::tests` (1 site, drives ~6 transitive supervisor test failures via shared fixture) All workspace tests green; clippy clean; fmt clean. Refs ai-gateway#345 ai-gateway#346

Integrate origin/main (commit 2c1d485 = post-PR-#326 / #348 plus #330 / #341 / #343 / #345 / #346) into this branch via `git merge --squash` to clear PR #344's lingering `mergeable: dirty` state. Conflict on `docs/quickstart/self-hosted.md` was a 3-way-merge-base artifact: base (3596c0a) read `- a reachable etcd instance`, main changed `a` → `A` (via #326), this branch additionally inserted the glossary link. Both changes are wanted; resolution per Umar's approved plan was `git checkout --ours`, which preserves the branch's self-hosted.md state (already integrates capital A + glossary link + first-time-build paragraph + keep-running framing). Other 4 overlapping doc files auto-merged cleanly (`bootstrap-config.md`, `core-concepts.md`, `first-model-first-key-first-request.md`, `openai-sdk.md`). Code files all auto-merged cleanly. Additional Copilot review (post-`167196a` cycle) addressed: - `docs/index.md:7` — change link display text from `[data-plane]` to `[data plane]` to match the canonical glossary term. The URL anchor `#data-plane` stays kebab-case (matches the glossary heading's auto-anchor); only the display text changes. Comment id 3271145422. - `docs/quickstart/openai-sdk.md:43` — change `All three steps below` to `All commands below`. The Install-the-SDK section has two command blocks (mkdir+cd, npm install), not three; the prior wording originated from a mental model (mkdir, cd, install) that doesn't match the typographic count of code blocks under the heading. Comment id 3271145458. Copilot's third comment on `docs/overview/core-concepts.md` Observability Exporter wording (id 3271145444) auto-resolves via this merge — main's #326 rewrite supersedes the branch's pre-#326 wording at that location ("ships per-request span telemetry… OTLP/HTTP-compatible backend…" replaces "Use this concept when documenting…"). No separate edit needed; the merge IS the fix.

…ickstart-polish Resolve PR #344's lingering mergeable: dirty state by linking the branch history to origin/main (2c1d485 = post-#326 / #348 / #330 / #341 / #343 / #345 / #346). The squash-merge commit landed earlier (e2af197) integrated main's content into the branch tree but did not link the histories, so GitHub's mergeable computation still saw the 3-way-merge-base artifact conflict on docs/quickstart/self-hosted.md (a vs A + the glossary link / "In another terminal" vs "Keep the gateway running" framing). This explicit merge commit ties the branch to main's history. Self-hosted.md conflict resolved by taking OUR side — the branch's edits already contain main's substantive changes (capital A, first-time-build paragraph) plus this PR's additions (glossary link, keep-running framing, YOUR_ADMIN_KEY note, config.yaml location anchor). The auto-merge of first-model-first-key-first-request.md duplicated the :::warning callout that was already integrated via the squash commit; removed the duplicate.

moonming · 2026-05-18T17:56:54Z

Summary

ai-gateway #345 (P2-A) added 11 long-tail OpenAI-adapter Provider::* variants + Hub registrations. The dashboard-side schema in schemas/resources/model.schema.json regenerated correctly. But there's a SEPARATELY-MAINTAINED hardcoded JSON schema in crates/aisix-core/src/models/schema.rs::model_schema() that backs aisix-etcd::loader's runtime validate_model check — and nobody updated it.

Result: cp-api admits a groq/mistral/etc. ProviderKey and Model (per #336 long-tail admission). The DP's aisix-etcd::loader reads the Model from etcd and validates against the OLD 6-value schema:

schema validation failed at `/provider`: "groq" is not one of ["openai","anthropic","google","deepseek","cohere","jina"]

Model rejected. Never appears in /v1/models. Customer's chat 404s on the alias.

Surfaced by AISIX-Cloud PR #366 e2e — first iteration of the D3.2 long-tail matrix. The groq Model creation succeeded at cp-api but the DP never saw it. Logs showed the schema rejection.

Fix

crates/aisix-core/src/models/schema.rs:120 — extend the provider enum from 6 to 17 values to match every Provider::as_str() output post-#345.

The existing negative-test model_unknown_provider_value_fails used "mistral" as the "unknown value" sentinel. Since mistral is now valid, renamed the fixture string to "this-is-not-a-provider-id" so the negative-test still pins the rejection class.

Why a separate schema even exists

aisix-etcd::loader::validate_and_parse runs schema validation BEFORE serde deserialization to fail-loud-and-skip on malformed entries (rather than crashing the DP). The schema is the gate; serde is the parser. Both must agree on the enum.

The dashboard-facing JSON schemas in schemas/resources/*.schema.json are GENERATED from the Rust types via cargo run --bin dump-schema (drift CI check enforces). The runtime validator's schema in schema.rs is hand-written and not gated by the drift check — a structural gap that this PR addresses with a behavioral fix only; a future refactor could fold the two sources together.

Test plan

cargo test -p aisix-core --lib schema — 52 tests pass
clippy clean
fmt clean
CI green
Audit (CLAUDE.md §7)
After merge + docker rebuild, AISIX-Cloud chore(fmt): restore rustfmt on aisix-etcd loader test (post-#362) #366 D3.2 batch 1 e2e exercises the unblocked path

References (per CLAUDE.md §7)

ai-gateway#345 (merged) — Provider enum + Hub register
aisix-etcd loader.rs:262 — the schema-rejection log path that surfaced the gap on PR chore(fmt): restore rustfmt on aisix-etcd loader test (post-#362) #366 CI

Refs api7/AISIX-Cloud#366

Summary by CodeRabbit

Improvements
- Expanded model provider validation to accept an extended set of provider identifiers during schema validation.
Tests
- Updated test cases to reflect the expanded provider validation rules, ensuring invalid providers continue to be properly rejected.

…ovider variants (#345 follow-up) ai-gateway #345 added 11 long-tail OpenAI-adapter Provider variants (groq, mistral, togetherai, fireworks-ai, perplexity, moonshotai, alibaba, zhipuai, baseten, huggingface, cerebras) to the Rust enum + Hub registrations + cp-api admission. The hardcoded JSON schema backing `aisix-etcd::loader`'s `validate_model` runtime check at `crates/aisix-core/src/models/schema.rs:120`, however, retained the original 6-value allowlist. The dashboard-side schema in `schemas/resources/model.schema.json` was regenerated correctly by #345's `dump-schema` step — that file IS now in sync with the enum. But the runtime validator uses a SEPARATELY-MAINTAINED hardcoded schema at line 112-200 of schema.rs that nobody updated. cp-api admission ✓, DP serde ✓, runtime schema validator ✗ — a customer's groq Model resource gets rejected at the etcd loader before deserialization with `schema validation failed at \`/provider\`: "groq" is not one of ["openai","anthropic","google","deepseek","cohere","jina"]`. This silently dropped EVERY long-tail Provider Model resource the dashboard wrote. The model never appears in `/v1/models` on the DP, and customer chat calls 404 on the alias. Surfaced by AISIX-Cloud PR #366 e2e (D3.2 batch 1) — the groq Model creation succeeds at cp-api but the DP never sees it. Logs showed the schema validation rejection. Fix `crates/aisix-core/src/models/schema.rs:120` — extend the `provider` enum to the full 17 values matching `Provider::as_str()` output across every variant: ```rust "provider": { "type": "string", "enum": [ "openai","anthropic","google","deepseek","cohere","jina", "groq","mistral","togetherai","fireworks-ai","perplexity", "moonshotai","alibaba","zhipuai","baseten","huggingface","cerebras" ] }, ``` The existing negative-test `model_unknown_provider_value_fails` used `"mistral"` as the "unknown value" sentinel; since mistral is now valid, the test would have stopped catching the rejection path. Renamed the test fixture value to `"this-is-not-a-provider-id"` so the negative-test still pins the rejection class. All 52 aisix-core schema tests pass; clippy clean; fmt clean. References (per CLAUDE.md §7) - ai-gateway#345 (merged) — Provider enum + Hub register additions - aisix-etcd loader.rs:262 — schema-rejection log path that surfaced the gap

coderabbitai · 2026-05-18T17:57:10Z

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Free

Run ID: 6608819e-114c-4760-8c03-7fa8bfd7b4c6

📥 Commits

Reviewing files that changed from the base of the PR and between 8d88b41 and ab76aa5.

📒 Files selected for processing (3)

crates/aisix-admin/src/lib.rs
crates/aisix-etcd/src/loader.rs
crates/aisix-etcd/src/supervisor.rs

✅ Files skipped from review due to trivial changes (1)

crates/aisix-etcd/src/supervisor.rs

📝 Walkthrough

Walkthrough

The PR expands the model JSON schema's provider field enum to accept a larger set of provider identifiers and updates tests/fixtures that previously used mistral to use a clearly invalid provider string so schema-rejection tests still fail.

Changes

Provider Field Allowlist Expansion

Layer / File(s)	Summary
Provider field schema and core validation test `crates/aisix-core/src/models/schema.rs`	The `provider` field enum in the model schema is expanded to accept a larger allowlist. The `model_unknown_provider_value_fails` test is updated to use an explicitly invalid provider string so the validation rejection path is still exercised.
Update tests and fixtures across crates `crates/aisix-admin/src/lib.rs`, `crates/aisix-etcd/src/loader.rs`, `crates/aisix-etcd/src/supervisor.rs`	Tests and JSON fixtures that previously used `provider: "mistral"` were changed to use `provider: "this-is-not-a-provider-id"` (or similar) to keep negative schema-validation tests failing under the expanded provider allowlist.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Note

🎁 Summarized by CodeRabbit Free

Your organization has reached its limit of developer seats under the Pro Plan. For new users, CodeRabbit will generate a high-level summary and a walkthrough for each pull request. For a comprehensive line-by-line review, please add seats to your subscription by visiting https://app.coderabbit.ai/login.If you believe this is a mistake and have available seats, please assign one to the pull request author through the subscription management page using the link above.

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

…rovider sentinel The negative-test fixtures across `aisix-admin` and `aisix-etcd` used `"mistral"` as the "unknown provider" sentinel — a vestige of the pre-#345 6-value allowlist. With ai-gateway#345 first-classing 11 long-tail variants (including mistral), those tests started asserting against a now-VALID provider and FAILED. Replaced every fixture occurrence with `"this-is-not-a-provider-id"`, which is intentionally NOT in the post-#345 17-value enum, keeping the schema-rejection class pinned without churn-tracking the live provider list. Files touched (post-#345 surgical sentinel rotation): - `crates/aisix-admin/src/lib.rs::create_model_with_invalid_provider_prefix_is_400_schema_error` - `crates/aisix-etcd/src/loader.rs::tests` (2 sites) - `crates/aisix-etcd/src/supervisor.rs::tests` (1 site, drives ~6 transitive supervisor test failures via shared fixture) All workspace tests green; clippy clean; fmt clean. Refs ai-gateway#345 ai-gateway#346

Integrate origin/main (commit 2c1d485 = post-PR-#326 / #348 plus #330 / #341 / #343 / #345 / #346) into this branch via `git merge --squash` to clear PR #344's lingering `mergeable: dirty` state. Conflict on `docs/quickstart/self-hosted.md` was a 3-way-merge-base artifact: base (3596c0a) read `- a reachable etcd instance`, main changed `a` → `A` (via #326), this branch additionally inserted the glossary link. Both changes are wanted; resolution per Umar's approved plan was `git checkout --ours`, which preserves the branch's self-hosted.md state (already integrates capital A + glossary link + first-time-build paragraph + keep-running framing). Other 4 overlapping doc files auto-merged cleanly (`bootstrap-config.md`, `core-concepts.md`, `first-model-first-key-first-request.md`, `openai-sdk.md`). Code files all auto-merged cleanly. Additional Copilot review (post-`167196a` cycle) addressed: - `docs/index.md:7` — change link display text from `[data-plane]` to `[data plane]` to match the canonical glossary term. The URL anchor `#data-plane` stays kebab-case (matches the glossary heading's auto-anchor); only the display text changes. Comment id 3271145422. - `docs/quickstart/openai-sdk.md:43` — change `All three steps below` to `All commands below`. The Install-the-SDK section has two command blocks (mkdir+cd, npm install), not three; the prior wording originated from a mental model (mkdir, cd, install) that doesn't match the typographic count of code blocks under the heading. Comment id 3271145458. Copilot's third comment on `docs/overview/core-concepts.md` Observability Exporter wording (id 3271145444) auto-resolves via this merge — main's #326 rewrite supersedes the branch's pre-#326 wording at that location ("ships per-request span telemetry… OTLP/HTTP-compatible backend…" replaces "Use this concept when documenting…"). No separate edit needed; the merge IS the fix.

…ickstart-polish Resolve PR #344's lingering mergeable: dirty state by linking the branch history to origin/main (2c1d485 = post-#326 / #348 / #330 / #341 / #343 / #345 / #346). The squash-merge commit landed earlier (e2af197) integrated main's content into the branch tree but did not link the histories, so GitHub's mergeable computation still saw the 3-way-merge-base artifact conflict on docs/quickstart/self-hosted.md (a vs A + the glossary link / "In another terminal" vs "Keep the gateway running" framing). This explicit merge commit ties the branch to main's history. Self-hosted.md conflict resolved by taking OUR side — the branch's edits already contain main's substantive changes (capital A, first-time-build paragraph) plus this PR's additions (glossary link, keep-running framing, YOUR_ADMIN_KEY note, config.yaml location anchor). The auto-merge of first-model-first-key-first-request.md duplicated the :::warning callout that was already integrated via the squash commit; removed the duplicate.

CopilotAI review requested due to automatic review settings May 18, 2026 17:56

Copilotstarted reviewing on behalf of moonming May 18, 2026 18:02 View session

moonming merged commit 71ea97e into mainMay 18, 2026
8 checks passed

jarvis9443 deleted the fix/runtime-schema-long-tail-providers branch June 25, 2026 06:26