fix(sessions): auto-compact on context overflow instead of failing the turn#340
Merged
Conversation
…e turn (#314) When the LLM provider rejects a request with a context overflow error, the session now triggers emergency compaction instead of failing the turn. Also hardens IsContextOverflowError detection: - Uses ProviderException.StatusCode (400) when available - Broader keyword matching for Anthropic ("prompt is too long") and vLLM ("token...exceed") in addition to Ollama/OpenAI patterns - 16 new tests covering detection across provider error formats The keyword matching is inherently brittle — there is no standard error format for context overflow across LLM providers. Tests document the known formats and catch regressions.
9d2f1f4 to
b1c1a12
Compare
2 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
When the LLM provider rejects a request with a context overflow error (e.g., "request (66540 tokens) exceeds the available context size (65536 tokens)"), the session now triggers emergency compaction instead of failing the turn.
Before:
LlmCallFailed→FailCurrentTurn()→ user sees error, session stuckAfter:
LlmCallFailed→ detect overflow → notify user →CompactionTriggered→ compact →ReadyUses the existing compaction pipeline (extractive reducer + observation summary). The user sees a message asking them to resend their last message after compaction completes.
Closes #314
Test plan
dotnet build— clean