alishahryar1--free-claude-code
Ali Khokhar
36bb282558
## Problem NVIDIA NIM can report context exhaustion as a 400 `BadRequestError` saying its derived `max_tokens` is negative. FCC treated that as a generic invalid request, so Claude Code could not recognize the smaller upstream context window, compact the conversation, and replay the interrupted turn. LM Studio also encoded Claude's recovery phrase inside provider code instead of reporting a protocol-neutral failure. Fixes #1198. ## Changes - Add one protocol-neutral `context_window_exceeded` execution failure with a non-retryable 400 contract. - Narrowly classify only NVIDIA NIM's negative derived-`max_tokens` signature while preserving the complete redacted provider diagnostic and request ID. - Let the Anthropic serializer alone add Claude's `prompt is too long` compaction trigger; OpenAI Responses keeps a standard invalid-request envelope. - Migrate LM Studio's existing context preflight to the same neutral semantic and document the ownership boundary. - Add provider, protocol, API, trace, near-miss, and real Claude compaction/replay coverage; bump FCC to 4.11.4. | Before | After | | --- | --- | | Context overflow appeared as an ordinary provider 400 and ended the Claude turn. | Claude receives a typed 400 with its recognized compaction trigger, compacts once, and replays the interrupted turn without an FCC or SDK retry loop. | <!-- greptile_comment --> <details open><summary><h3>Greptile Summary</h3></summary> This PR adds protocol-neutral recovery from provider context-window exhaustion. The main changes are: - Classify NVIDIA NIM negative derived-`max_tokens` errors as context-window failures. - Move Claude's compaction trigger into the Anthropic serializer. - Migrate LM Studio context preflight to the neutral failure type. - Preserve the standard OpenAI Responses invalid-request envelope. - Add provider, protocol, API, trace, and recovery tests. - Bump the package version to 4.11.4. </details> <h3>Confidence Score: 5/5</h3> This looks safe to merge. - No blocking issues found in the changed code. - The new failure kind is covered by both protocol mappings. - Provider classification remains narrow and non-retryable. - The protocol-specific compaction phrase stays at the Anthropic boundary. <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 narrow pytest command from /home/user/repo without live provider credentials or services, and observed a clean test run with all tests passing. <a href="https://app.greptile.com/trex/runs/15073403/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/nvidia_nim/client.py | Adds narrow context-window classification for nested and top-level NVIDIA NIM error bodies. | | src/free_claude_code/providers/lmstudio/client.py | Migrates context-budget preflight failures to the canonical context-window failure. | | src/free_claude_code/core/anthropic/errors.py | Adds Anthropic mapping and injects Claude's compaction phrase at the wire boundary. | | src/free_claude_code/core/openai_responses/errors.py | Maps context exhaustion to the standard OpenAI invalid-request error. | | src/free_claude_code/providers/failure_policy.py | Adds the canonical non-retryable status-400 context-window failure factory. | | src/free_claude_code/core/failures.py | Adds the protocol-neutral context-window failure category. | </details> <details open><summary><h3>Flowchart</h3></summary> <a href="#gh-light-mode-only"> ```mermaid %%{init: {'theme': 'neutral'}}%% flowchart TD A[Provider detects context exhaustion] --> B[Context-window ExecutionFailure] B --> C{Protocol adapter} C -->|Anthropic Messages| D[400 invalid_request_error] D --> E[Add prompt is too long trigger] E --> F[Claude compacts and replays] C -->|OpenAI Responses| G[400 invalid_request_error] G --> H[Keep neutral provider message] ``` </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[Provider detects context exhaustion] --> B[Context-window ExecutionFailure] B --> C{Protocol adapter} C -->|Anthropic Messages| D[400 invalid_request_error] D --> E[Add prompt is too long trigger] E --> F[Claude compacts and replays] C -->|OpenAI Responses| G[400 invalid_request_error] G --> H[Keep neutral provider message] ``` </a> </details> <sub>Reviews (1): Last reviewed commit: ["Normalize provider context overflow for ..."](https://github.com/alishahryar1/free-claude-code/commit/29c89a4c1011d986284e9d163855174d1d433293) | [Re-trigger Greptile](https://app.greptile.com/api/retrigger?id=45603631)</sub> <!-- /greptile_comment -->