main
7 次代码提交
| 作者 | SHA1 | 备注 | 提交日期 | |
|---|---|---|---|---|
 Ali Khokhar
|
65a342ede4 |
Honor reasoning tiers on numeric-budget providers (#1150)
## Problem FCC preserved named reasoning effort but numeric-budget providers received no intensity unless a client supplied an exact token budget. NIM and llama.cpp therefore treated Low through Max like the same provider default. ## Changes | Before | After | | --- | --- | | Named effort had no FCC-owned numeric meaning. | FCC maps Low, Medium, High, X-High, and Max to 512, 1,024, 2,048, 4,096, and 8,192 tokens. | | NIM received only thinking booleans for named effort. | [NIM](https://docs.nvidia.com/nim/large-language-models/1.15.0/thinking-budget-control.html) receives thinking booleans plus the mapped `reasoning_budget`, retaining its existing retry without rejected budget control. | | llama.cpp forwarded only exact client budgets. | [llama.cpp](https://github.com/ggml-org/llama.cpp/blob/master/tools/server/README.md) receives the mapped `thinking_budget_tokens` value. | | Named, boolean, and provider-default adapters shared no explicit numeric contract. | Named adapters keep words, boolean adapters keep on/off, and only numeric-budget adapters consume the FCC scale. | | Documentation prohibited every named-effort budget conversion. | Documentation defines the product scale, exact-budget precedence, and model-independent ownership boundary. | | Version 4.8.0 exposed tiers that collapsed on numeric providers. | Version 4.8.1 completes the tiers; all five local CI checks pass with 2,383 tests passed and 40 skipped. | <!-- greptile_comment --> <details open><summary><h3>Greptile Summary</h3></summary> This PR gives named reasoning tiers numeric budgets for providers that require token counts. The main changes are: - Adds one shared tier-to-token scale with exact-budget precedence. - Sends mapped budgets to NVIDIA NIM and llama.cpp. - Removes conflicting client-supplied NIM budget fields. - Adds focused provider and policy tests. - Updates documentation and bumps the package to 4.8.1. </details> <h3>Confidence Score: 5/5</h3> This looks safe to merge. The NIM fix removes both conflicting budget locations while preserving unrelated nested options. No blocking issue remains in the changed paths. <details><summary><h3><a href="https://www.greptile.com/trex"><img alt="T-Rex" src="https://greptile-static-assets.s3.amazonaws.com/trex/trex_green.svg" height="20" align="absmiddle"></a> T-Rex Logs</h3></summary> **What T-Rex did** - T-Rex ran a pre-change focused validation against the budget logic and observed 16 failures and 2 passes, establishing the baseline. - T-Rex ran a post-change focused validation (head run) and confirmed all 18 cases passed, including the mocked NIM retry paths. - T-Rex executed the full focused validation after the change and produced a verbose log showing 91/91 passing. <a href="https://app.greptile.com/trex/runs/14800553/artifacts"><picture><source media="(prefers-color-scheme: dark)" srcset="https://greptile-static-assets.s3.amazonaws.com/badges/ViewAllArtifactsDark.svg?v=4"><source media="(prefers-color-scheme: light)" srcset="https://greptile-static-assets.s3.amazonaws.com/badges/ViewAllArtifacts.svg?v=4"><img alt="View all artifacts" src="https://greptile-static-assets.s3.amazonaws.com/badges/ViewAllArtifacts.svg?v=4"></picture></a> <sub><a href="https://www.greptile.com/trex"><img alt="T-Rex" src="https://greptile-static-assets.s3.amazonaws.com/trex/trex_green.svg" height="14" align="absmiddle"></a> Ran code and verified through T-Rex</sub> </details> <details open><summary><h3>Important Files Changed</h3></summary> | Filename | Overview | |----------|----------| | src/free_claude_code/core/reasoning.py | Adds the shared effort-to-token scale and preserves exact client budgets as the higher-priority value. | | src/free_claude_code/providers/nvidia_nim/request_options.py | Removes top-level and nested client budgets before inserting one policy-derived NIM budget. | | src/free_claude_code/providers/openai_chat/reasoning.py | Extends llama.cpp request encoding to send mapped named-effort budgets. | </details> <sub>Reviews (2): Last reviewed commit: ["fix: canonicalize NIM reasoning budgets"](https://github.com/alishahryar1/free-claude-code/commit/368f1a88d3f1404fcab2365703cca89f1077f1ea) | [Re-trigger Greptile](https://app.greptile.com/api/retrigger?id=44998377)</sub> <!-- /greptile_comment --> |
||
|
Ali Khokhar
|
6455c63e1d |
Make reasoning policy provider-neutral and client-aware (#1148)
## Problem FCC reduced reasoning to global and route booleans, mixing client intent, configuration, provider wire capabilities, output visibility, and history replay. That discarded named client efforts, encouraged model-name checks, and made provider behavior inconsistent. ## Changes | Before | After | | --- | --- | | Admin exposed global and route thinking toggles. | Admin exposes **Off**, **From client**, **Low**, **Medium**, **High**, **X-High**, and **Max**; Fable, Opus, Sonnet, and Haiku also expose **Inherit**. | | Request intent was repeatedly reduced to a boolean across routing and providers. | The application boundary resolves one immutable `ReasoningPolicy` with independent control, named effort, and exact positive token budget. | | Provider adapters could infer reasoning behavior from upstream model names or versions. | Provider profiles translate only documented provider-wide wire capabilities; architecture and contributor rules prohibit model-specific reasoning branches. | | Gateway reasoning controls were ad hoc. | [OpenRouter](https://openrouter.ai/docs/guides/best-practices/reasoning-tokens) and [Vercel AI Gateway](https://vercel.com/docs/ai-gateway/models-and-providers) use documented reasoning objects, including exact budgets where representable. | | Named effort forwarding was inconsistent or absent. | [Gemini](https://ai.google.dev/gemini-api/docs/openai), [Ollama](https://docs.ollama.com/api/openai-compatibility), [LM Studio](https://lmstudio.ai/changelog/lmstudio-v0.4.8), [Fireworks](https://docs.fireworks.ai/guides/querying-text-models/reasoning), [Cohere](https://docs.cohere.com/docs/compatibility-api), [Wafer](https://docs.wafer.ai/serverless/api-reference), [Groq](https://console.groq.com/docs/reasoning), [Cerebras](https://inference-docs.cerebras.ai/capabilities/reasoning), [SambaNova](https://docs.sambanova.ai/docs/api-reference/chat-completions/create-chat-based-completion), and [Mistral](https://docs.mistral.ai/studio-api/conversations/reasoning) receive their documented named vocabularies with explicit provider-owned downgrades. | | Boolean thinking controls were mixed into shared conversion. | [DeepSeek](https://api-docs.deepseek.com/guides/thinking_mode/), [Kimi](https://platform.kimi.ai/docs/guide/use-kimi-k2-thinking-model), [Z.ai](https://docs.z.ai/guides/capabilities/thinking-mode), [Cloudflare Workers AI](https://developers.cloudflare.com/changelog/post/2026-04-20-kimi-k2-6-workers-ai/), and [NVIDIA NIM](https://docs.nvidia.com/nim/large-language-models/1.15.0/thinking-budget-control.html) use provider-owned thinking-object or chat-template controls. | | Effort names and output limits could become fabricated reasoning budgets. | Exact budgets remain exact and are forwarded only through documented fields for OpenRouter, Fireworks, LM Studio, NIM, and [llama.cpp](https://github.com/ggml-org/llama.cpp/blob/master/tools/server/README.md); named efforts and output limits are never converted into token budgets. | | New-turn reasoning and prior-turn replay shared one switch. | Every profile independently declares native reasoning replay, `<think>` tag replay, provider-specific replay, or no replay; **Off** suppresses new reasoning output without corrupting required history. | | Providers without a stable generic compute control received guessed controls. | [MiniMax](https://platform.minimax.io/docs/api-reference/text-openai-api) requests split output only, while [GitHub Models](https://docs.github.com/en/rest/models/inference), [Hugging Face Inference Providers](https://huggingface.co/docs/inference-providers/en/tasks/chat-completion), Codestral, and OpenCode keep provider defaults and use only their explicit replay profile. | | OpenAI Responses effort became a lossy Anthropic thinking boolean. | Responses preserves `reasoning.effort` through `output_config`, then resolves it through the same application policy as Messages without inventing a budget. | | Legacy booleans remained the persisted contract. | FCC-owned dotenv files migrate to typed `REASONING_*` values, explicit env files receive an actionable warning, documentation describes the ownership boundary, and the package advances to 4.8.0. | | Reasoning behavior was covered by scattered boolean assertions. | New policy, routing, encoder, provider, Admin, migration, Responses, and smoke contracts pass all five local CI checks: 2,368 tests passed, 40 skipped; 92 smoke tests collect and both live config migration checks pass. | <!-- greptile_comment --> <details open><summary><h3>Greptile Summary</h3></summary> This PR makes reasoning policy client-aware and independent of provider model names. The main changes are: - Adds one immutable reasoning policy resolved at the application boundary. - Adds typed root and route reasoning settings with Admin UI support. - Moves wire controls and history replay behavior into provider profiles. - Migrates owned dotenv files from legacy thinking booleans. - Expands provider, routing, migration, API, and smoke coverage. </details> <h3>Confidence Score: 5/5</h3> This looks safe to merge. No blocking issues found in the changed code. <details><summary><h3><a href="https://www.greptile.com/trex"><img alt="T-Rex" src="https://greptile-static-assets.s3.amazonaws.com/trex/trex_green.svg" height="20" align="absmiddle"></a> T-Rex Logs</h3></summary> **What T-Rex did** - Ran the contract-validation test suite with the specified test modules, and the tests reported 78 passed in 1.53s with exit code 0. - Reviewed the complete captured output artifact reasoning-contract-02-after.log to verify the final test outcomes and successful contract validation. <a href="https://app.greptile.com/trex/runs/14792858/artifacts"><picture><source media="(prefers-color-scheme: dark)" srcset="https://greptile-static-assets.s3.amazonaws.com/badges/ViewAllArtifactsDark.svg?v=4"><source media="(prefers-color-scheme: light)" srcset="https://greptile-static-assets.s3.amazonaws.com/badges/ViewAllArtifacts.svg?v=4"><img alt="View all artifacts" src="https://greptile-static-assets.s3.amazonaws.com/badges/ViewAllArtifacts.svg?v=4"></picture></a> <sub><a href="https://www.greptile.com/trex"><img alt="T-Rex" src="https://greptile-static-assets.s3.amazonaws.com/trex/trex_green.svg" height="14" align="absmiddle"></a> Ran code and verified through T-Rex</sub> </details> <details open><summary><h3>Important Files Changed</h3></summary> | Filename | Overview | |----------|----------| | src/free_claude_code/config/env_migrations.py | Migrates legacy reasoning booleans in owned dotenv files and warns for explicit environment files. | | src/free_claude_code/application/reasoning.py | Resolves client controls and configured preferences into one provider-neutral reasoning policy. | | src/free_claude_code/application/routing.py | Carries route-level reasoning preferences into request-scoped policy resolution. | | src/free_claude_code/providers/openai_chat/reasoning.py | Provides shared provider encoders for reasoning controls and replay behavior. | </details> <sub>Reviews (2): Last reviewed commit: ["chore: release reasoning controls as 4.8..."](https://github.com/alishahryar1/free-claude-code/commit/9d4be767f7dbdca5709474012f43dcdc6f4347e3) | [Re-trigger Greptile](https://app.greptile.com/api/retrigger?id=44984039)</sub> <!-- /greptile_comment --> |
||
|
Ali Khokhar
|
4f56a6aa17 |
Add Fable as a first-class Claude routing tier (#1099)
## Problem Claude Code now sends `claude-fable-5` for the Fable alias, but FCC treated it as an unrecognized model and collapsed it into the global fallback route. Users could not map Fable traffic or reasoning behavior independently. Fixes #1097. ## Changes | Before | After | | --- | --- | | Fable requests inherited `MODEL` and `ENABLE_MODEL_THINKING`. | Fable requests use `MODEL_FABLE` and `ENABLE_FABLE_THINKING` when configured, otherwise inherit the existing defaults. | | `/v1/models` omitted Claude Fable 5. | `/v1/models` advertises the canonical `claude-fable-5` identifier. | | Admin, documentation, validation, and smoke contracts described three Claude tiers. | Admin, documentation, validation, and smoke contracts describe Fable alongside Opus, Sonnet, and Haiku. | | The package version was `4.3.1`. | The package version is `4.4.0`. | <!-- greptile_comment --> <details open><summary><h3>Greptile Summary</h3></summary> This PR adds Fable as a Claude routing tier. The main changes are: - `MODEL_FABLE` and `ENABLE_FABLE_THINKING` settings. - Fable routing and thinking resolution in `ModelRouter`. - `claude-fable-5` in the model catalog. - Admin, docs, smoke, and test coverage updates. - Package version bump to `4.4.0`. </details> <h3>Confidence Score: 4/5</h3> The changed routing path needs a fix for direct provider model ids containing `fable`. Fable settings, validation, admin fields, and model listing are consistent with the existing tier patterns. Blank Fable settings inherit the existing defaults. Direct provider model ids can receive the Fable thinking override when their model name contains `fable`. src/free_claude_code/application/routing.py <details><summary><h3><a href="https://www.greptile.com/trex"><img alt="T-Rex" src="https://greptile-static-assets.s3.amazonaws.com/trex/trex_green.svg" height="20" align="absmiddle"></a> T-Rex Logs</h3></summary> **What T-Rex did** - Reproduced the Fable thinking overmatch by running a focused Python repro that disables global thinking and enables Fable thinking, then resolves sambanova/my-fable-ensemble-v2. - The repro confirmed direct routing preserved provider\_id=sambanova, provider\_model=my-fable-ensemble-v2, and provider\_model\_ref=sambanova/my-fable-ensemble-v2, with resolved\_thinking\_enabled and thinking\_enabled both true. - Ran the fable-tier validation pytest, which finished with exit code 0 and 177 tests passed. - Ran the runtime probe to exercise the API model list, Settings env parsing, and ModelRouter paths, and observed a 200 OK on GET /v1/models, with the catalog item id claude-fable-5 and the expected Fable vs global-default routing behavior. - Generated the probe source file used to exercise the API and Settings paths, enabling repeatable validation without real provider credentials. <a href="https://app.greptile.com/trex/runs/14275379/artifacts"><picture><source media="(prefers-color-scheme: dark)" srcset="https://greptile-static-assets.s3.amazonaws.com/badges/ViewAllArtifactsDark.svg?v=4"><source media="(prefers-color-scheme: light)" srcset="https://greptile-static-assets.s3.amazonaws.com/badges/ViewAllArtifacts.svg?v=4"><img alt="View all artifacts" src="https://greptile-static-assets.s3.amazonaws.com/badges/ViewAllArtifacts.svg?v=4"></picture></a> <sub><a href="https://www.greptile.com/trex"><img alt="T-Rex" src="https://greptile-static-assets.s3.amazonaws.com/trex/trex_green.svg" height="14" align="absmiddle"></a> Ran code and verified through T-Rex</sub> </details> <details open><summary><h3>Important Files Changed</h3></summary> | Filename | Overview | |----------|----------| | src/free_claude_code/application/routing.py | Adds Fable model and thinking branches; the thinking branch can also match unrelated direct provider model ids containing `fable`. | | src/free_claude_code/config/settings.py | Adds optional Fable model and thinking settings with blank-env inheritance and provider/model validation. | | src/free_claude_code/config/model_refs.py | Includes Fable in configured chat model reference collection and dedupe. | | src/free_claude_code/config/admin/manifest.py | Adds Fable model and thinking controls to the admin manifest. | | src/free_claude_code/api/model_catalog.py | Adds `claude-fable-5` to the advertised Claude model aliases. | | pyproject.toml | Bumps the package version to `4.4.0`. | | uv.lock | Updates the editable package version to match `pyproject.toml`. | </details> <details open><summary><h3>Flowchart</h3></summary> <a href="#gh-light-mode-only"> ```mermaid %%{init: {'theme': 'neutral'}}%% flowchart TD A[Incoming model name] --> B{Direct provider or gateway id?} B -- yes --> C[Use provider/model directly] C --> D[Resolve thinking from provider model string] B -- no --> E{Claude tier match} E -- Fable --> F[MODEL_FABLE or MODEL] E -- Opus/Sonnet/Haiku --> G[Tier override or MODEL] E -- None --> H[MODEL] D --> I[Provider request] F --> I G --> I H --> I ``` </a> <a href="#gh-dark-mode-only"> ```mermaid %%{init: {'theme': 'base', 'themeVariables': {"darkMode": true, "background": "#0d1117", "primaryColor": "#21262d", "primaryTextColor": "#e6edf3", "primaryBorderColor": "#8b949e", "lineColor": "#8b949e", "textColor": "#e6edf3", "edgeLabelBackground": "#161b22", "actorBkg": "#21262d", "actorBorder": "#8b949e", "actorTextColor": "#e6edf3", "actorLineColor": "#8b949e", "signalColor": "#8b949e", "signalTextColor": "#e6edf3", "noteBkgColor": "#373320", "noteBorderColor": "#d4a72c", "noteTextColor": "#f0e6c0", "labelBoxBkgColor": "#21262d", "labelBoxBorderColor": "#8b949e", "labelTextColor": "#e6edf3", "loopTextColor": "#e6edf3", "activationBkgColor": "#30363d", "activationBorderColor": "#8b949e"}}}%% flowchart TD A[Incoming model name] --> B{Direct provider or gateway id?} B -- yes --> C[Use provider/model directly] C --> D[Resolve thinking from provider model string] B -- no --> E{Claude tier match} E -- Fable --> F[MODEL_FABLE or MODEL] E -- Opus/Sonnet/Haiku --> G[Tier override or MODEL] E -- None --> H[MODEL] D --> I[Provider request] F --> I G --> I H --> I ``` </a> </details> <a href="https://app.greptile.com/api/ide/codex?prompt=IMPORTANT%3A%20Work%20in%20the%20repository%20%22alishahryar1%2Ffree-claude-code%22%20on%20the%20existing%20branch%20%22ali%2Fadd-fable-routing-tier%22.%20Checkout%20that%20branch%20%E2%80%94%20do%20NOT%20create%20a%20new%20branch%20or%20open%20a%20new%20PR.%20Push%20your%20changes%20to%20%22ali%2Fadd-fable-routing-tier%22.%0A%0AFix%20the%20following%201%20code%20review%20issue.%20Work%20through%20them%20one%20at%20a%20time%2C%20proposing%20concise%20fixes.%0A%0A---%0A%0A%23%23%23%20Issue%201%20of%201%0Asrc%2Ffree_claude_code%2Fapplication%2Frouting.py%3A134-135%0A**Fable%20Thinking%20Overmatches%20Models**%0A%0AWhen%20a%20direct%20provider%20request%20uses%20a%20model%20id%20like%20%60sambanova%2Fmy-fable-ensemble-v2%60%2C%20the%20direct%20route%20bypasses%20tier%20remapping%20but%20still%20calls%20%60_resolve_thinking%28%29%60%20with%20the%20provider%20model%20string.%20With%20%60ENABLE_FABLE_THINKING%60%20set%2C%20this%20substring%20check%20applies%20Fable%20thinking%20behavior%20to%20an%20unrelated%20provider%20model%2C%20changing%20the%20outgoing%20request%20shape%20just%20because%20the%20model%20id%20contains%20%60fable%60.%0A%0A&repo=alishahryar1%2Ffree-claude-code&pr=1099&platform=github"><picture><source media="(prefers-color-scheme: dark)" srcset="https://greptile-static-assets.s3.amazonaws.com/badges/FixAllInCodexDark.svg?v=6"><source media="(prefers-color-scheme: light)" srcset="https://greptile-static-assets.s3.amazonaws.com/badges/FixAllInCodex.svg?v=6"><img alt="Fix All in Codex" src="https://greptile-static-assets.s3.amazonaws.com/badges/FixAllInCodex.svg?v=6"></picture></a> <sub>Reviews (1): Last reviewed commit: ["Add Fable as a first-class routing tier"](https://github.com/alishahryar1/free-claude-code/commit/0705840c65511fd85c74fd2c62ba8ea97afe7c12) | [Re-trigger Greptile](https://app.greptile.com/api/retrigger?id=43892323)</sub> > Greptile also left **1 inline comment** on this PR. <!-- /greptile_comment --> |
||
|
Ali Khokhar
|
3a7e0ccf7a |
Remove the native Anthropic provider transport (#1067)
## Problem Ollama and llama.cpp still used a parallel native Anthropic transport after the other providers moved to OpenAI Chat. That kept duplicate request, SSE, recovery, model-list, and server-tool policy machinery alive. ## Changes | Before | After | | --- | --- | | Ollama and llama.cpp streamed through provider-specific Anthropic `/messages` adapters. | Ollama and llama.cpp use the shared OpenAI Chat transport. | | Native request serialization, SSE normalization, error mapping, and recovery remained beside the OpenAI path. | Native-only machinery is removed and all providers share one transport lifecycle. | | Routed models carried a capability object solely to permit native server-tool passthrough. | Routing carries only route decisions; FCC handles forced server tools locally and rejects lossy passthrough. | | Ollama discovery used a separate `/api/tags` parser and rejected `/v1` configuration. | Ollama discovery uses `/v1/models` and accepts either root or `/v1` base URLs. | | Obsolete native tests and a compatibility facade kept deleted internals represented. | Tests cover the shared transport and real Ollama product path without compatibility shims. | <!-- greptile_comment --> <details open><summary><h3>Greptile Summary</h3></summary> This PR removes the native Anthropic transport path for local providers. The main changes are: - Ollama and llama.cpp now use the shared OpenAI Chat transport. - Local provider base URLs are normalized to the OpenAI-compatible `/v1` API root. - Ollama discovery now uses the OpenAI-compatible model-listing path. - Native Anthropic transport code and server-tool passthrough capability metadata were removed. - Tests and smoke coverage were updated for the shared transport path. </details> <h3>Confidence Score: 5/5</h3> This looks safe to merge. No blocking issues found in the changed code. None. <details><summary><h3><a href="https://www.greptile.com/trex"><img alt="T-Rex" src="https://greptile-static-assets.s3.amazonaws.com/trex/trex_green.svg" height="20" align="absmiddle"></a> T-Rex Logs</h3></summary> **What T-Rex did** - Ran the targeted local provider pytest slice and observed exit code 0. - Executed the generated runtime harness to emulate the provider HTTP interactions and capture a request trace. - Validated the request trace showed two GET /v1/models calls authorized as Bearer ollama for root and /v1 base URL configurations, and a POST /v1/chat/completions authorized as Bearer llamacpp with streaming OpenAI chat JSON payload. - Confirmed the exact generated harness script used for the runtime proof is the harness file referenced in the artifacts. <a href="https://app.greptile.com/trex/runs/14121848/artifacts"><picture><source media="(prefers-color-scheme: dark)" srcset="https://greptile-static-assets.s3.amazonaws.com/badges/ViewAllArtifactsDark.svg?v=4"><source media="(prefers-color-scheme: light)" srcset="https://greptile-static-assets.s3.amazonaws.com/badges/ViewAllArtifacts.svg?v=4"><img alt="View all artifacts" src="https://greptile-static-assets.s3.amazonaws.com/badges/ViewAllArtifacts.svg?v=4"></picture></a> <sub><a href="https://www.greptile.com/trex"><img alt="T-Rex" src="https://greptile-static-assets.s3.amazonaws.com/trex/trex_green.svg" height="14" align="absmiddle"></a> Ran code and verified through T-Rex</sub> </details> <details open><summary><h3>Important Files Changed</h3></summary> | Filename | Overview | |----------|----------| | src/free_claude_code/providers/transports/openai_chat/base_url.py | Adds a helper that normalizes local OpenAI-compatible server roots to `/v1`. | | src/free_claude_code/providers/llamacpp/client.py | Moves llama.cpp to the shared OpenAI Chat transport with local base URL normalization. | | src/free_claude_code/providers/ollama/client.py | Moves Ollama to the shared OpenAI Chat transport with local base URL normalization. | | src/free_claude_code/api/handlers/messages.py | Applies server-tool rejection through the shared request policy instead of provider passthrough metadata. | | src/free_claude_code/application/routing.py | Removes provider capability metadata from routed model results. | </details> <!-- greptile_failed_comments --> <h3>Comments Outside Diff (1)</h3> 1. `src/free_claude_code/api/handlers/messages.py`, line 261-267 ([link](https://github.com/alishahryar1/free-claude-code/blob/3c7ff176da46560c4d27b3846dca1ab1c7db561c/src/free_claude_code/api/handlers/messages.py#L261-L267)) <a href="#"><img alt="P2" src="https://greptile-static-assets.s3.amazonaws.com/badges/p2.svg?v=9" align="top"></a> **Native Server Tools Always Reject** With the passthrough capability check removed, Ollama and llama.cpp requests that previously used their native Anthropic transport for `web_search` or `web_fetch` are rejected before provider execution. The default `ENABLE_WEB_SERVER_TOOLS=false` now makes forced server-tool requests return an invalid-request error instead of reaching the local provider path that used to support them. <a href="https://app.greptile.com/api/ide/codex?prompt=IMPORTANT%3A%20Work%20in%20the%20repository%20%22alishahryar1%2Ffree-claude-code%22%20on%20the%20existing%20branch%20%22ali%2Fremove-native-anthropic-transport%22.%20Checkout%20that%20branch%20%E2%80%94%20do%20NOT%20create%20a%20new%20branch%20or%20open%20a%20new%20PR.%20Push%20your%20changes%20to%20%22ali%2Fremove-native-anthropic-transport%22.%0A%0AThis%20is%20a%20comment%20left%20during%20a%20code%20review.%0APath%3A%20src%2Ffree_claude_code%2Fapi%2Fhandlers%2Fmessages.py%0ALine%3A%20261-267%0A%0AComment%3A%0A**Native%20Server%20Tools%20Always%20Reject**%0A%0AWith%20the%20passthrough%20capability%20check%20removed%2C%20Ollama%20and%20llama.cpp%20requests%20that%20previously%20used%20their%20native%20Anthropic%20transport%20for%20%60web_search%60%20or%20%60web_fetch%60%20are%20rejected%20before%20provider%20execution.%20The%20default%20%60ENABLE_WEB_SERVER_TOOLS%3Dfalse%60%20now%20makes%20forced%20server-tool%20requests%20return%20an%20invalid-request%20error%20instead%20of%20reaching%20the%20local%20provider%20path%20that%20used%20to%20support%20them.%0A%0AHow%20can%20I%20resolve%20this%3F%20If%20you%20propose%20a%20fix%2C%20please%20make%20it%20concise.&repo=alishahryar1%2Ffree-claude-code&pr=1067&platform=github"><picture><source media="(prefers-color-scheme: dark)" srcset="https://greptile-static-assets.s3.amazonaws.com/badges/FixInCodexDark.svg?v=6"><source media="(prefers-color-scheme: light)" srcset="https://greptile-static-assets.s3.amazonaws.com/badges/FixInCodex.svg?v=6"><img alt="Fix in Codex" src="https://greptile-static-assets.s3.amazonaws.com/badges/FixInCodex.svg?v=6"></picture></a> <!-- /greptile_failed_comments --> <sub>Reviews (2): Last reviewed commit: ["Normalize local OpenAI v1 base URLs"](https://github.com/alishahryar1/free-claude-code/commit/2355eac247a6e411f89a781f46e784636ced98d6) | [Re-trigger Greptile](https://app.greptile.com/api/retrigger?id=43572841)</sub> <!-- /greptile_comment --> |
||
|
Ali Khokhar
|
5ffa47fbc3 |
Make response stream lifetimes explicit (#1060)
## Problem Client disconnects and response-start send failures could abandon a prefetched provider stream and its generation lease. Re-yielding iterators and response-proxy middleware left no owner that closed the complete body chain before runtime release. ## Changes | Before | After | | --- | --- | | Starlette body iteration indirectly owned stream cleanup and lease release. | One FCC streaming response surrounds the real ASGI send, closes the body transitively, then releases the lease exactly once. | | The prefetched first-frame generator could not close its tail before replay began. | An explicit closeable replay iterator owns the prefetched tail in every commit state. | | Tracing, execution, Responses conversion, and native transport transforms re-yielded inputs without closing them. | Every retained transform closes its direct input; redundant transport wrappers are removed while provider construction failures remain deferred. | | Function-style correlation middleware proxied and canceled streaming responses. | Pure ASGI correlation spans the complete stream, preserves request headers and log context, and keeps the catch-all 500 fallback correlated. | | Repeated cancellation could interrupt pre-start and post-start cleanup. | Shielded completion tasks finish body closure before release and then restore caller cancellation. | | The package version was 3.5.5. | The package version is 3.5.6; full CI passes with 2,162 tests and stable live API/provider/disconnect/client smoke passes 63 scenarios. | <!-- greptile_comment --> <details open><summary><h3>Greptile Summary</h3></summary> This PR makes streaming response ownership explicit across the API path. The main changes are: - Adds a managed streaming response that closes the body chain before releasing provider resources. - Adds a prefetched replay iterator for first-frame commit handling. - Moves request correlation to pure ASGI middleware for full-stream context. - Propagates direct-input closure through execution, tracing, Responses conversion, and provider transports. - Bumps the package version and updates tests for stream cleanup behavior. </details> <h3>Confidence Score: 5/5</h3> This looks safe to merge. No blocking issues found in the changed code. None. <details><summary><h3><a href="https://www.greptile.com/trex"><img alt="T-Rex" src="https://greptile-static-assets.s3.amazonaws.com/trex/trex_green.svg" height="20" align="absmiddle"></a> T-Rex Logs</h3></summary> **What T-Rex did** - Validated the execution environment by reviewing the environment proof log, confirming uv 0.11.28, CPython 3.14.0, a repo-local virtual environment, and exit code 0. - Verified that the requested test command was executed, based on the test proof log. - Confirmed the test run completed successfully with 91 tests passing in 3.63 seconds, as shown in the test proof log. <a href="https://app.greptile.com/trex/runs/14099258/artifacts"><picture><source media="(prefers-color-scheme: dark)" srcset="https://greptile-static-assets.s3.amazonaws.com/badges/ViewAllArtifactsDark.svg?v=4"><source media="(prefers-color-scheme: light)" srcset="https://greptile-static-assets.s3.amazonaws.com/badges/ViewAllArtifacts.svg?v=4"><img alt="View all artifacts" src="https://greptile-static-assets.s3.amazonaws.com/badges/ViewAllArtifacts.svg?v=4"></picture></a> <sub><a href="https://www.greptile.com/trex"><img alt="T-Rex" src="https://greptile-static-assets.s3.amazonaws.com/trex/trex_green.svg" height="14" align="absmiddle"></a> Ran code and verified through T-Rex</sub> </details> <details open><summary><h3>Important Files Changed</h3></summary> | Filename | Overview | |----------|----------| | src/free_claude_code/api/response_streams.py | Adds the managed response owner, first-frame replay iterator, and shielded cleanup flow. | | src/free_claude_code/api/request_ids.py | Adds pure ASGI request correlation and response-start header injection. | | src/free_claude_code/core/trace.py | Adds shared stream input closure tracing and closes traced inputs on exit. | | src/free_claude_code/application/execution.py | Closes provider stream iterators from the executor wrapper when streaming ends. | | src/free_claude_code/providers/transports/anthropic_messages/transport.py | Returns provider runner streams directly and closes layered SSE iterators explicitly. | </details> <sub>Reviews (2): Last reviewed commit: ["Make response stream lifetimes explicit"](https://github.com/alishahryar1/free-claude-code/commit/cb698c62c08924d5f80a1cea7dbd19c0b8af26a2) | [Re-trigger Greptile](https://app.greptile.com/api/retrigger?id=43527620)</sub> <!-- /greptile_comment --> |
||
|
Ali Khokhar
|
701a394697 |
Make provider capabilities semantic and typed (#1047)
## Problem Provider metadata mixed descriptive strings and transport-family names with product policy. The Messages API had to know provider implementation details to decide server-tool behavior, while locality and thinking metadata had ambiguous owners. ## Changes | Before | After | | --- | --- | | Provider descriptors exposed transport names and untyped capability strings. | Provider descriptors expose an immutable semantic value containing only locality and server-tool passthrough. | | The Messages handler derived provider sets from transport families. | `ResolvedModel` carries catalog capabilities and the handler asks the resolved semantic policy. | | Discovery, Admin, and smoke selection inferred locality from credentials or string membership. | Discovery, Admin, and smoke selection read the catalog-owned `local` capability. | | Provider-wide thinking strings competed with discovered model metadata. | `ProviderModelInfo.supports_thinking` remains the sole per-model thinking authority. | | Invalid mapped providers could become raw catalog lookup failures during the migration. | Routing and provider construction share the canonical typed unknown-provider error. | | LM Studio smoke contracts described a native Anthropic transport. | LM Studio smoke contracts describe its OpenAI-chat-backed Messages path. | | Tests encoded transport identities as product behavior. | Tests prove both directions of capability-versus-identity independence while preserving existing wire behavior and diagnostics. | <!-- greptile_comment --> <details open><summary><h3>Greptile Summary</h3></summary> This PR moves provider policy from transport names to typed catalog capabilities. The main changes are: - Added immutable provider capability values for locality and server-tool passthrough. - Routed provider capabilities through `ResolvedModel` for Messages API policy. - Updated admin status, discovery, smoke config, docs, and tests to read semantic capabilities. - Centralized unknown-provider diagnostics for routing and provider construction. - Bumped the package version and lockfile to `3.4.22`. </details> <h3>Confidence Score: 5/5</h3> This looks safe to merge. No blocking issues found in the changed code. None. <details><summary><h3><a href="https://www.greptile.com/trex"><img alt="T-Rex" src="https://greptile-static-assets.s3.amazonaws.com/trex/trex_green.svg" height="20" align="absmiddle"></a> T-Rex Logs</h3></summary> **What T-Rex did** - The Pytest run for the provider capabilities changes was executed and completed successfully, as shown by the log's EXIT\_CODE: 0. - The Ruff linter run for the provider capabilities changes was executed and completed successfully, as shown by the log's EXIT\_CODE: 0. - The focused suite exercised the changed code paths via repository tests rather than relying solely on static inspection. <a href="https://app.greptile.com/trex/runs/14074560/artifacts"><picture><source media="(prefers-color-scheme: dark)" srcset="https://greptile-static-assets.s3.amazonaws.com/badges/ViewAllArtifactsDark.svg?v=4"><source media="(prefers-color-scheme: light)" srcset="https://greptile-static-assets.s3.amazonaws.com/badges/ViewAllArtifacts.svg?v=4"><img alt="View all artifacts" src="https://greptile-static-assets.s3.amazonaws.com/badges/ViewAllArtifacts.svg?v=4"></picture></a> <sub><a href="https://www.greptile.com/trex"><img alt="T-Rex" src="https://greptile-static-assets.s3.amazonaws.com/trex/trex_green.svg" height="14" align="absmiddle"></a> Ran code and verified through T-Rex</sub> </details> <details open><summary><h3>Important Files Changed</h3></summary> | Filename | Overview | |----------|----------| | src/free_claude_code/config/provider_catalog.py | Replaces transport and string capability metadata with typed immutable provider capabilities. | | src/free_claude_code/application/routing.py | Adds catalog capabilities to resolved models and uses the shared unknown-provider error. | | src/free_claude_code/api/handlers/messages.py | Uses resolved semantic capabilities to decide server-tool passthrough. | | src/free_claude_code/config/admin/status.py | Classifies local providers through the new catalog capability. | | src/free_claude_code/providers/runtime/discovery.py | Uses typed locality to decide which providers are discovered only when referenced. | | src/free_claude_code/application/errors.py | Adds a shared constructor for canonical unknown-provider messages. | </details> <sub>Reviews (1): Last reviewed commit: ["Make provider capabilities semantic and ..."](https://github.com/alishahryar1/free-claude-code/commit/b2732ef74d61c632ccae332a19aaa655f45a96e7) | [Re-trigger Greptile](https://app.greptile.com/api/retrigger?id=43475337)</sub> <!-- /greptile_comment --> |
||
|
Ali Khokhar
|
f8c21a48f2 |
Introduce a typed application boundary for provider execution (#1045)
## Problem The HTTP adapter owned model routing, provider execution, and runtime-facing contracts, so API handlers depended on provider implementation types. Provider preflight also discovered private request builders dynamically, obscuring the boundary that must fail before streaming begins. ## Changes | Before | After | | --- | --- | | `api/` owned model routing and shared provider execution. | `application/` owns routing and a settings-independent `ProviderExecutor`. | | API handlers accepted `BaseProvider` callbacks. | API handlers consume the narrow structural `ProviderPort`. | | `BaseProvider` discovered `_build_request_body` dynamically. | Both transport families implement explicit abstract preflight, with LM Studio composing context validation. | | Request leases, task control, and provider model metadata had adapter/provider owners. | Application-owned ports and immutable values define those cross-package contracts. | | Boundary direction was implicit. | Architecture contracts and documentation enforce the final dependency direction. | | Package version was `3.4.19`. | Package version is `3.4.20`, with the lockfile updated. | | Coverage followed the old module layout. | Deterministic boundary/preflight regressions and live Messages/Responses smokes cover the new shape. | <!-- greptile_comment --> <details open><summary><h3>Greptile Summary</h3></summary> This PR adds a typed application boundary for provider execution. The main changes are: - New `application` package for routing, execution, ports, and model metadata. - API handlers now call application-owned routing and provider execution. - Provider preflight is now explicit on the transport families. - Runtime API composition now uses a task-control port for `/stop`. - Import-boundary tests, smoke references, docs, version, and lockfile were updated. </details> <h3>Confidence Score: 5/5</h3> This looks safe to merge. No blocking issues found in the changed code. None. <details><summary><h3><a href="https://www.greptile.com/trex"><img alt="T-Rex" src="https://greptile-static-assets.s3.amazonaws.com/trex/trex_green.svg" height="20" align="absmiddle"></a> T-Rex Logs</h3></summary> **What T-Rex did** - Ran a deterministic pytest run for the provider boundary preflight, which completed with exit code 0 and 102 tests passing in 4.77 seconds. - Launched the environment presence check as part of the preflight, which completed with exit code 0 and confirmed OPENCODE\_API\_KEY=\[REDACTED\] matched. - Attempted the live provider smoke test, which completed with exit code 0 and 2 tests skipped due to incomplete smoke configuration. <a href="https://app.greptile.com/trex/runs/14067905/artifacts"><picture><source media="(prefers-color-scheme: dark)" srcset="https://greptile-static-assets.s3.amazonaws.com/badges/ViewAllArtifactsDark.svg?v=4"><source media="(prefers-color-scheme: light)" srcset="https://greptile-static-assets.s3.amazonaws.com/badges/ViewAllArtifacts.svg?v=4"><img alt="View all artifacts" src="https://greptile-static-assets.s3.amazonaws.com/badges/ViewAllArtifacts.svg?v=4"></picture></a> <sub><a href="https://www.greptile.com/trex"><img alt="T-Rex" src="https://greptile-static-assets.s3.amazonaws.com/trex/trex_green.svg" height="14" align="absmiddle"></a> Ran code and verified through T-Rex</sub> </details> <details open><summary><h3>Important Files Changed</h3></summary> | Filename | Overview | |----------|----------| | src/free_claude_code/application/execution.py | Moves shared provider execution into the application layer and keeps eager preflight before token counting and streaming. | | src/free_claude_code/application/ports.py | Adds structural provider, request-runtime, and task-control protocols used across the new boundary. | | src/free_claude_code/api/routes.py | Updates route composition to use the application provider resolver and task-control stop path. | | src/free_claude_code/providers/base.py | Makes provider preflight explicit by requiring subclasses or transport bases to implement it. | | src/free_claude_code/providers/transports/openai_chat/transport.py | Adds OpenAI-chat preflight through the same request-body builder used by streaming. | | src/free_claude_code/providers/transports/anthropic_messages/transport.py | Adds native Messages preflight through the native request-body builder. | | src/free_claude_code/providers/model_listing.py | Keeps provider model-list parsing while moving `ProviderModelInfo` ownership to the application layer. | | src/free_claude_code/runtime/bootstrap.py | Passes the runtime object through the new `tasks` service slot. | | tests/contracts/test_import_boundaries.py | Extends import-boundary tests for the new application package. | </details> <details open><summary><h3>Flowchart</h3></summary> <a href="#gh-light-mode-only"> ```mermaid %%{init: {'theme': 'neutral'}}%% flowchart LR API[api handlers and routes] --> Routing[application.routing] API --> Executor[application.execution] API --> Ports[application.ports] Executor --> ProviderPort[ProviderPort] ProviderPort --> Preflight[preflight_stream] ProviderPort --> Stream[stream_response] Runtime[runtime bootstrap and provider manager] --> Ports Providers[providers] --> Metadata[application.model_metadata] Executor --> Core[core anthropic and trace] Routing --> Config[config settings and model refs] ``` </a> <a href="#gh-dark-mode-only"> ```mermaid %%{init: {'theme': 'base', 'themeVariables': {"darkMode": true, "background": "#0d1117", "primaryColor": "#21262d", "primaryTextColor": "#e6edf3", "primaryBorderColor": "#8b949e", "lineColor": "#8b949e", "textColor": "#e6edf3", "edgeLabelBackground": "#161b22", "actorBkg": "#21262d", "actorBorder": "#8b949e", "actorTextColor": "#e6edf3", "actorLineColor": "#8b949e", "signalColor": "#8b949e", "signalTextColor": "#e6edf3", "noteBkgColor": "#373320", "noteBorderColor": "#d4a72c", "noteTextColor": "#f0e6c0", "labelBoxBkgColor": "#21262d", "labelBoxBorderColor": "#8b949e", "labelTextColor": "#e6edf3", "loopTextColor": "#e6edf3", "activationBkgColor": "#30363d", "activationBorderColor": "#8b949e"}}}%% flowchart LR API[api handlers and routes] --> Routing[application.routing] API --> Executor[application.execution] API --> Ports[application.ports] Executor --> ProviderPort[ProviderPort] ProviderPort --> Preflight[preflight_stream] ProviderPort --> Stream[stream_response] Runtime[runtime bootstrap and provider manager] --> Ports Providers[providers] --> Metadata[application.model_metadata] Executor --> Core[core anthropic and trace] Routing --> Config[config settings and model refs] ``` </a> </details> <sub>Reviews (1): Last reviewed commit: ["Introduce typed application boundary"](https://github.com/alishahryar1/free-claude-code/commit/4cdcf97231c812c6f568ca3d74af7ce759d7f2dc) | [Re-trigger Greptile](https://app.greptile.com/api/retrigger?id=43462788)</sub> <!-- /greptile_comment --> |