Skip to content

fix(context): resolve context windows per provider/model across status and run accounting#27439

Closed
paul-tian wants to merge 1 commit intoopenclaw:mainfrom
paul-tian:pr/context-window
Closed

fix(context): resolve context windows per provider/model across status and run accounting#27439
paul-tian wants to merge 1 commit intoopenclaw:mainfrom
paul-tian:pr/context-window

Conversation

@paul-tian
Copy link
Copy Markdown

@paul-tian paul-tian commented Feb 26, 2026 •
edited
Loading

Summary

Describe the problem and fix in 2–5 bullets:

  • Problem: Currently the /status is showing context window based on model only
  • Why it matters: same model from different providers may have different ctx window limit
  • What changed: now the ctx window will calculate base on the chosen provider/model
  • What did NOT change (scope boundary): other parts

Change Type (select all)

  • Bug fix
  • Feature
  • Refactor
  • Docs
  • Security hardening
  • Chore/infra

Scope (select all touched areas)

  • Gateway / orchestration
  • Skills / tool execution
  • Auth / tokens
  • Memory / storage
  • Integrations
  • API / contracts
  • UI / DX
  • CI/CD / infra

Linked Issue/PR

User-visible / Behavior Changes

List user-visible changes (including defaults/config). /status correctly shows the ctx window limit base on chosen model
If none, write None.

Security Impact (required)

  • New permissions/capabilities? (Yes/No) No
  • Secrets/tokens handling changed? (Yes/No) No
  • New/changed network calls? (Yes/No) No
  • Command/tool execution surface changed? (Yes/No) No
  • Data access scope changed? (Yes/No) No
  • If any Yes, explain risk + mitigation:

Repro + Verification

Environment

  • OS: macOS 26.3
  • Runtime/container: direct deployment
  • Model/provider: openai/gpt-5.3-codex and openai-codex/gpt-5.3-codex
  • Integration/channel (if any): Telegram
  • Relevant config (redacted):

Steps

  1. Switch between model openai/gpt-5.3-codex and openai-codex/gpt-5.3-codex
  2. check ctx window using /status

Expected

  • the ctx window update correspondingly

Actual

  • the ctx window update correspondingly

Evidence

Attach at least one:

  • Failing test/log before + passing after
  • Trace/log snippets
  • Screenshot/recording
  • Perf numbers (if relevant)

api
oauth

Human Verification (required)

What you personally verified (not just CI), and how:

  • Verified scenarios: direct deployment, Telegram channel
  • Edge cases checked: N/A
  • What you did not verify: everything else

Compatibility / Migration

  • Backward compatible? (Yes/No) Yes
  • Config/env changes? (Yes/No) No
  • Migration needed? (Yes/No) No
  • If yes, exact upgrade steps:

Failure Recovery (if this breaks)

  • How to disable/revert this change quickly:
  • Files/config to restore:
  • Known bad symptoms reviewers should watch for:

Risks and Mitigations

List only real risks for this PR. Add/remove entries as needed. If none, write None.

  • Risk: None

@openclaw-barnacle openclaw-barnacle bot added gateway Gateway runtime agents Agent runtime and tooling size: M labels Feb 26, 2026
@paul-tian paul-tian marked this pull request as ready for review February 26, 2026 11:18
@greptile-apps
Copy link
Copy Markdown
Contributor

greptile-apps bot commented Feb 26, 2026

Greptile Summary

This PR correctly implements provider-scoped context window resolution, fixing the issue where /status showed incorrect context windows when the same model was available from different providers with different limits.

  • Core change: Added provider-aware cache keys (provider/model) alongside legacy keys (just model) in src/agents/context.ts
  • Lookup strategy: Prioritizes provider-scoped lookups first, falls back to legacy keys for backward compatibility
  • Fail-safe behavior: When multiple providers expose the same model ID, keeps the smaller window for legacy lookups to prevent token overestimation
  • Comprehensive update: All call sites (agent-runner.ts, directive-handling.persist.ts, followup-runner.ts, status.ts, run.ts, session-utils.ts) now pass provider information to resolveContextTokensForModel()
  • Bonus fix: Corrected followup-runner.ts:241 to persist the actual providerUsed instead of incorrectly passing fallbackProvider
  • Status refresh: The /status command now re-resolves context tokens after parsing transcript metadata, ensuring accurate display when provider/model are updated from logs

Confidence Score: 5/5

  • This PR is safe to merge with minimal risk
  • The implementation is straightforward and well-structured. All changes maintain backward compatibility through dual-key lookups (scoped and legacy), edge cases are handled properly (empty strings, missing providers), and the test mock is updated. The code includes proper normalization, fail-safe min logic for collisions, and a bonus bug fix. No logical errors or potential runtime issues identified.
  • No files require special attention

Last reviewed commit: 5eb5cd1

@AaronL725 AaronL725 mentioned this pull request Feb 26, 2026
Closed
@paul-tian paul-tian force-pushed the pr/context-window branch from 5eb5cd1 to 9cacf94 Compare March 3, 2026 12:28
@WhiteGiverMa WhiteGiverMa mentioned this pull request Mar 5, 2026
Closed
@paul-tian paul-tian closed this Mar 30, 2026
@paul-tian paul-tian deleted the pr/context-window branch March 30, 2026 04:20
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

agents Agent runtime and tooling gateway Gateway runtime size: M

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

1 participant

@paul-tian