fix(kimi-code): correct maxTokens defaults and add server-side override by myk1yt · Pull Request #1216 · Zoo-Code-Org/Zoo-Code

myk1yt · 2026-08-10T06:05:03Z

Summary

Kimi Code OAuth provider displayed incorrect Max Output values (32,768) for all models. According to Kimi's official documentation, the correct default for kimi-k3 is 131,072.

Root Cause

Three issues in the codebase:

kimiCodeDefaultModelInfo.maxTokens was hardcoded to 32_768 instead of 131_072
The Zod schema for /models API response didn't include max_tokens field (silently stripped)
mapKimiCodeModel() overrode contextWindow from server but not maxTokens

Changes (3 files, +47/-1)

File	Change
`packages/types/src/providers/kimi-code.ts`	Update `maxTokens` to `131_072`, add `kimiCodeModelDefaults` for per-model fallback
`src/api/providers/fetchers/kimi-code.ts`	Add `max_tokens` to Zod schema, override in `mapKimiCodeModel()`
`src/api/providers/fetchers/__tests__/kimi-code.spec.ts`	Add 4 tests for maxTokens override and fallback

Fallback Chain

maxTokens = server.max_tokens
         ?? kimiCodeModelDefaults[model.id].maxTokens
         ?? kimiCodeDefaultModelInfo.maxTokens (131,072)

This ensures automatic updates when the server provides max_tokens, with per-model and global fallbacks.

Verification

Vitest: 11/11 passed (7 existing + 4 new)
ESLint: clean
TypeScript: clean

Summary by CodeRabbit

New Features
- Added smarter tool-call handling based on model and provider capabilities.
- Improved support for single-call and parallel-call workflows, including safer handling of incomplete or duplicate calls.
- Expanded MiMo compatibility, including flexible endpoint configuration and accurate usage-based cost reporting.
- Increased Kimi Code model token limits with model-specific defaults.
Bug Fixes
- Improved recovery from unsupported request options and malformed tool-call responses.
- Fixed Mistral streaming cost calculations and usage reporting.
Tests
- Added comprehensive coverage for tool-call behavior, provider compatibility, and cost tracking.

# Conflicts: # src/core/tools/error-interception/StructuralValidator.ts

# Conflicts: # src/core/assistant-message/__tests__/NativeToolCallParser.spec.ts # src/core/assistant-message/__tests__/presentAssistantMessage-parser-dedup.integration.spec.ts # src/core/assistant-message/presentAssistantMessage.ts

# Conflicts: # src/core/assistant-message/__tests__/presentAssistantMessage-parser-dedup.integration.spec.ts

…capabilities

…ception refs from backup

MimoHandler was passing raw tool schemas to the API without the strict mode conversion that all other OpenAI-compatible providers use. This caused tool call errors due to missing required/strict fields. - Call this.convertToolsForOpenAI(tools) instead of raw assignment - Adds strict: true, required properties, additionalProperties: false

…de arg

An id-less argument-continuation chunk belongs to the most recent id chunk seen at its index. When a provider reuses index 0 with a NEW id (a disguised second parallel call), the new call's id chunk was dropped but its id-less argument fragments were still kept and concatenated into the FIRST call's accumulator, corrupting its JSON. Track dropped indexes in filterToFirstToolCall state and drop subsequent id-less fragments for those indexes. Also rewrite the function docblock, which referenced a non-existent error-interception retry loop.

The parseErrors/parseFailures docblocks claimed presentAssistantMessage routes recorded failures to an INVALID_JSON_ARGUMENTS error-interception pattern. No such routing exists on this codebase; describe the actual lifecycle (consumed via the consume* APIs, cleared on new API request). Comment-only change, no behavior difference.

parseErrors/parseFailures static maps accumulated an entry per malformed tool call and were never cleared in production (the consume* APIs have no production callers), slowly leaking for the extension-host lifetime. Add NativeToolCallParser.clearParseFailures() and call it in Task.recursivelyMakeClineRequests alongside clearAllStreamingToolCalls()/ clearRawChunkState(), where other per-stream state is reset. The consume* APIs keep working for tests.

MiMo sends tools through convertToolsForOpenAI(), which attaches a strict flag to every function tool. An OpenAI-compatible endpoint that doesn't support structured outputs rejects the request with a 400 and the turn fails outright. Mirror the existing parallel_tool_calls fallback: detect schema-rejection errors narrowly (400 status plus a mention of strict/additionalProperties in a tools context, so unrelated 400s like MiMo's missing-reasoning_content rejection are not retried) and retry once with the original schemas and no strict flag.

…ecov/patch

…ssions

… yield

…ypes)

…1132) CI failure: E2E Tests (Mocked) failed with '404 No fixture matched' because provider-cost.test.ts calls startNewTask with probe tag 'provider-cost-e2e' but no fixture existed.

…stubs and custom endpoints

…x partial check

Fixes Zoo-Code-Org#1215 - Update kimiCodeDefaultModelInfo.maxTokens from 32768 to 131072 - Add kimiCodeModelDefaults for per-model fallback values - Add max_tokens to kimiCodeModelSchema for server response parsing - Override maxTokens in mapKimiCodeModel: server > model-default > global-default - Add 4 new tests for maxTokens override and fallback behavior

coderabbitai · 2026-08-10T06:05:39Z

Caution

Review failed

The pull request is closed.

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 397c164d-f08d-4950-9b04-dc7b0a2e1f67

📥 Commits

Reviewing files that changed from the base of the PR and between fcc507c and dc31fca.

📒 Files selected for processing (29)

apps/vscode-e2e/fixtures/provider-cost.json
apps/vscode-e2e/src/suite/provider-cost.test.ts
codecov.yml
packages/telemetry/src/TelemetryService.ts
packages/types/src/model.ts
packages/types/src/provider-settings.ts
packages/types/src/providers/kimi-code.ts
packages/types/src/providers/mimo.ts
packages/types/src/telemetry.ts
progress.txt
src/api/index.ts
src/api/providers/__tests__/mimo.spec.ts
src/api/providers/__tests__/mistral.spec.ts
src/api/providers/fetchers/__tests__/kimi-code.spec.ts
src/api/providers/fetchers/kimi-code.ts
src/api/providers/mimo.ts
src/api/providers/mistral.ts
src/core/assistant-message/NativeToolCallParser.ts
src/core/assistant-message/ToolCallRetentionPolicy.ts
src/core/assistant-message/__tests__/NativeToolCallParser.spec.ts
src/core/assistant-message/__tests__/ToolCallRetentionPolicy-telemetry.spec.ts
src/core/assistant-message/__tests__/ToolCallRetentionPolicy.spec.ts
src/core/prompts/tools/native-tools/execute_command.ts
src/core/task/Task.ts
src/core/task/__tests__/ghost-quarantine.spec.ts
src/core/task/__tests__/tool-call-policy.spec.ts
src/core/tools/ExecuteCommandTool.ts
src/eslint-suppressions.json
src/shared/tools.ts

📝 Walkthrough

Walkthrough

The change adds model-aware tool-call policies, MiMo single-call handling, ghost-call quarantine, classified parser failures, metadata-only telemetry, Kimi Code token defaults, provider cost validation, and related type and configuration updates.

Changes

Tool-call control

Layer / File(s)	Summary
Policy and telemetry contracts `packages/types/src/model.ts`, `packages/types/src/providers/mimo.ts`, `packages/types/src/telemetry.ts`, `src/api/index.ts`, `packages/telemetry/src/TelemetryService.ts`	Model capabilities and provider metadata now resolve tool-call generation and enforcement policies. New telemetry events record policy metadata and enforcement counts without raw tool data.
Parser failures and call retention `src/core/assistant-message/NativeToolCallParser.ts`, `src/core/assistant-message/ToolCallRetentionPolicy.ts`, `src/core/assistant-message/__tests__/*`	Parser failures are classified and consumable through typed and legacy APIs. Retention helpers classify empty ghosts, select executable calls, and emit metadata-only enforcement telemetry.
MiMo tool-call handling `src/api/providers/mimo.ts`, `src/api/providers/__tests__/mimo.spec.ts`	MiMo requests use typed parameters, retry unsupported fields, and filter streamed responses to the first tool call. Tests cover request forwarding, retries, streaming, malformed calls, and duplicate IDs.
Task policy integration and ghost quarantine `src/core/task/Task.ts`, `src/core/task/__tests__/*`, `src/core/prompts/tools/native-tools/execute_command.ts`	Task requests use resolved parallel-call settings. Streaming, legacy, and unterminated calls discard provably empty ghosts, clean state, and emit telemetry. Tool documentation reflects the `cwd` contract.

Provider metadata and cost handling

Layer / File(s)	Summary
Kimi Code model limits `packages/types/src/providers/kimi-code.ts`, `src/api/providers/fetchers/kimi-code.ts`, `src/api/providers/fetchers/__tests__/kimi-code.spec.ts`	Kimi Code model mapping now prioritizes server `max_tokens`, then per-model defaults, then the global default. The global default is 131,072 tokens.
Provider usage cost validation `src/api/providers/mistral.ts`, `src/api/providers/__tests__/mistral.spec.ts`, `apps/vscode-e2e/...`	Mistral usage events now include calculated cost. The MiMo E2E test validates streamed usage and persisted pricing using a local SSE fixture.
Provider and tool contracts `packages/types/src/provider-settings.ts`, `src/core/tools/ExecuteCommandTool.ts`, `src/shared/tools.ts`, `codecov.yml`	MiMo base URLs accept valid URLs. Execute-command timeouts no longer accept `null`. Default patch coverage is informational.

Estimated code review effort: 5 (Critical) | ~120 minutes

Possibly related issues

feat: Add Error Interception Middleware for guided AI self-correcting tool errors #1000 — Both changes modify native tool-call error handling in NativeToolCallParser and Task, although this PR adds retention and ghost-call enforcement.

Possibly related PRs

Zoo-Code-Org/Zoo-Code#210 — Shares MiMo streaming tool-call handling and tool-call ID sanitization changes.
Zoo-Code-Org/Zoo-Code#945 — Shares Kimi Code model metadata and token-default changes.
Zoo-Code-Org/Zoo-Code#1130 — Shares the MiMo tool-call policy, telemetry, parser, retention, and task integration changes.

Suggested labels: awaiting-review

Suggested reviewers: taltas

✨ Finishing Touches

🧪 Generate unit tests (beta)

Create PR with unit tests

Warning

There were issues while running some tools. Please review the errors and either fix the tool's configuration or disable the tool if it's a critical failure.

🔧 ESLint

If the error stems from missing dependencies, add them to the package.json file. For unrecoverable errors (e.g., due to private dependencies), disable the tool in the CodeRabbit configuration.

apps/vscode-e2e/src/suite/provider-cost.test.ts

ESLint skipped: missing config or dependency (missing-dependency). The ESLint configuration references a package that is not available in the sandbox.

packages/telemetry/src/TelemetryService.ts

ESLint skipped: missing config or dependency (missing-dependency). The ESLint configuration references a package that is not available in the sandbox.

packages/types/src/model.ts

ESLint skipped: missing config or dependency (missing-dependency). The ESLint configuration references a package that is not available in the sandbox.

22 others

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

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

Warning

⚠️ This pull request shows signs of AI-generated slop (description_diff_mismatch). It has been flagged by CodeRabbit slop detection and should be reviewed carefully.

myk1yt · 2026-08-10T06:09:13Z

Closing this PR because it includes unrelated changes from the fork. Reopening a clean PR with only kimi-code changes.

codecov · 2026-08-10T06:10:12Z

Codecov Report

❌ Patch coverage is 68.68687% with 62 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/core/task/Task.ts	12.24%	41 Missing and 2 partials ⚠️
src/core/assistant-message/NativeToolCallParser.ts	71.05%	7 Missing and 4 partials ⚠️
src/api/providers/mimo.ts	93.65%	0 Missing and 4 partials ⚠️
packages/telemetry/src/TelemetryService.ts	0.00%	2 Missing ⚠️
src/api/providers/mistral.ts	33.33%	0 Missing and 2 partials ⚠️

📢 Thoughts on this report? Let us know!

Zoo (VP) and others added 30 commits August 7, 2026 04:11

feat: add model-level tool-call capability and policy resolution

76d3837

feat: wire MiMo provider controls and tighten argument normalization

c290d77

# Conflicts: # src/core/tools/error-interception/StructuralValidator.ts

feat: add tool-call policy telemetry events

7fba2f0

# Conflicts: # src/core/assistant-message/__tests__/presentAssistantMessage-parser-dedup.integration.spec.ts

fix: resolve no-explicit-any lint errors in mimo and telemetry files

307baa4

fix: preserve parallel behavior for known providers without explicit …

45b99c7

…capabilities

fix: port NativeToolParseFailure infrastructure and clean error-inter…

47deb83

…ception refs from backup

fix: port cleaned mimo.ts provider from backup to match spec types

5f21220

fix(mimo): suppress parallel tool calls at provider stream level

e07abd9

fix(mimo): pass openAiToolStrictMode setting to convertToolsForOpenAI

3cd9242

fix(mimo): pass tools to convertToolsForOpenAI without extra strictMo…

a43d4d9

…de arg

test(b12): add mimo error retry and ghost quarantine coverage for cod…

2a6842e

…ecov/patch

fix(test): replace 'as any' with typed assertion in mistral.spec.ts

b68d77b

fix(b17): re-encode mimo.spec.ts to UTF-8, remove stale eslint suppre…

d52f9aa

…ssions

fix(b17): strip BOM from mimo.spec.ts, add totalCost to mistral usage…

341c861

… yield

test(b17): add 4 coverage tests for mimo.ts edge cases

50f84a5

fix(b17): remove stale mimo.spec.ts eslint-suppression entry (0 any t…

61efc90

…ypes)

fix: prune stale eslint-suppressions.json entries

0b229dd

fix: make codecov/patch informational for PRs with large new code

af018ae

chore: remove temp file progress.txt

7767c23

Merge branch 'main' into pr/b17-provider-cost-v2

f125045

fix: prune unused eslint suppressions

3c4b293

test(e2e): add provider cost suite

d2070d6

fix(test): add aimock fixture for provider-cost e2e (PR Zoo-Code-Org#…

17d2c14

…1132) CI failure: E2E Tests (Mocked) failed with '404 No fixture matched' because provider-cost.test.ts calls startNewTask with probe tag 'provider-cost-e2e' but no fixture existed.

fix(types,e2e): allow any valid URL for mimoBaseUrl to support local …

712bb66

…stubs and custom endpoints

Zoo (VP) added 2 commits August 8, 2026 16:52

fix(vscode-e2e): emit attempt_completion in provider-cost stub and fi…

59fcd0e

…x partial check

myk1yt requested review from JamesRobert20, edelauna, hannesrudolph, navedmerchant and taltas as code owners August 10, 2026 06:05

Merge branch 'main' into fix/kimi-code

dc31fca

myk1yt closed this Aug 10, 2026

Navigation Menu

Uh oh!

fix(kimi-code): correct maxTokens defaults and add server-side override - #1216

fix(kimi-code): correct maxTokens defaults and add server-side override#1216
myk1yt wants to merge 33 commits into
Zoo-Code-Org:mainfrom
myk1yt:fix/kimi-code

myk1yt commented Aug 10, 2026 •

edited by coderabbitai Bot

Loading

Uh oh!

coderabbitai Bot commented Aug 10, 2026 •

edited

Loading

Review failed

Walkthrough

Changes

Uh oh!

myk1yt commented Aug 10, 2026

Uh oh!

codecov Bot commented Aug 10, 2026 •

edited

Loading

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Uh oh!

Conversation

myk1yt commented Aug 10, 2026 • edited by coderabbitai Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Root Cause

Changes (3 files, +47/-1)

Fallback Chain

Verification

Summary by CodeRabbit

Uh oh!

coderabbitai Bot commented Aug 10, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Review failed

Walkthrough

Changes

Uh oh!

myk1yt commented Aug 10, 2026

Uh oh!

codecov Bot commented Aug 10, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

myk1yt commented Aug 10, 2026 •

edited by coderabbitai Bot

Loading

coderabbitai Bot commented Aug 10, 2026 •

edited

Loading

codecov Bot commented Aug 10, 2026 •

edited

Loading