PR #74289: fix: preserve K2.6 reasoning payload fields to prevent incomplete terminal replies

• Repository: openclaw/openclaw
• Author: jsdevninja
• State: open | merged: False
• Link: https://github.com/openclaw/openclaw/pull/74289

Description (problem / solution / changelog)

Summary

• Problem: Fireworks Kimi K2.6 (kimi-k2p6 / kimi-k2.6-*) could end turns with Agent couldn't generate a response, which was classified as an incomplete terminal response in fallback handling.
• Why it matters: This causes failed agent replies and contributes to degraded operator experience (including status/probe confusion reported in the issue).
• What changed: In extensions/fireworks/stream.ts, K2.6 models now keep reasoning* payload fields while still forcing thinking: { type: "disabled" }; tests were updated in extensions/fireworks/stream.test.ts.
• What did NOT change (scope boundary): No gateway probe logic, no model-fallback classifier logic, no auth/token flow, no non-Fireworks provider behavior.

Change Type (select all)

• Bug fix
• Feature
• Refactor required for the fix
• 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

Closes #74205

• This PR fixes a bug or regression

Root Cause (if applicable)

• Root cause: Fireworks Kimi wrapper deleted reasoning, reasoning_effort, and reasoningEffort unconditionally while disabling thinking; for K2.6, that can remove content needed for a visible terminal reply.
• Missing detection / guardrail: No K2.6-specific regression test asserting reasoning field preservation when thinking is disabled.
• Contributing context (if known): K2.6 behavior differs from older Kimi variants already covered by existing assumptions/tests.

Regression Test Plan (if applicable)

For bug fixes or regressions, name the smallest reliable test coverage that should catch this. Otherwise write N/A.

• Coverage level that should have caught this:

  • Unit test
  • Seam / integration test
  • End-to-end test
  • Existing coverage already sufficient

• Target test or file: extensions/fireworks/stream.test.ts

• Scenario the test should lock in: K2.6 payload keeps reasoning* fields while still setting thinking: { type: "disabled" }; K2.5 behavior remains unchanged (reasoning stripped).

• Why this is the smallest reliable guardrail: The bug is in provider payload patching logic, which is fully deterministically verifiable at unit level.

• Existing test that already covers this (if any): Existing tests covered K2.5 stripping and thinking-disable behavior, but not K2.6 preservation.

• If no new test is added, why not: N/A (new test added).

User-visible / Behavior Changes

• Fireworks Kimi K2.6 requests are less likely to fail with an incomplete terminal response due to over-stripped payload fields.
• No config/default changes.

Diagram (if applicable)

Before:
[user prompt] -> [Fireworks K2.6 payload patch strips reasoning* + disables thinking]
-> [model returns no visible terminal text]
-> [incomplete terminal response path]

After:
[user prompt] -> [Fireworks K2.6 payload patch keeps reasoning* + disables thinking]
-> [visible terminal reply preserved]
-> [normal reply path]

Security Impact (required)

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

Repro + Verification

Environment

• OS:
• Runtime/container:
• Model/provider:
• Integration/channel (if any):
• Relevant config (redacted):

Steps

1. Configure OpenClaw default model to Fireworks Kimi K2.6.
2. Trigger an embedded agent turn.
3. Observe reply behavior before/after patch.

Expected

-Agent returns a visible terminal reply without incomplete terminal response fallback.

Actual

• Before: can hit incomplete terminal response path.
• After: payload shaping no longer strips K2.6 reasoning fields; targeted tests pass for new behavior.

Evidence

Attach at least one:

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

Human Verification (required)

• Verified scenarios: Ran pnpm test extensions/fireworks/stream.test.ts and pnpm test extensions/fireworks/index.test.ts; both pass with new K2.6 guard.
• Edge cases checked: K2.5 still strips reasoning fields; K2.6 keeps reasoning fields; non-target providers/models still unaffected by wrapper selection logic.
• What you did not verify: Full end-to-end live provider run against Fireworks API and gateway status behavior on Ubuntu/systemd.

Review Conversations

• I replied to or resolved every bot review conversation I addressed in this PR.
• I left unresolved only the conversations that still need reviewer or maintainer judgment.

If a bot review conversation is addressed by this PR, resolve that conversation yourself. Do not leave bot review conversation cleanup for maintainers.

Compatibility / Migration

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

Risks and Mitigations

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

• Risk:
  • Mitigation:

Changed files

• extensions/fireworks/stream.test.ts (modified, +33/-1)
• extensions/fireworks/stream.ts (modified, +27/-4)

PR #74327: fix #74205: [Bug]: kimi/k2p6 returns incomplete terminal response and gateway status probe times out

• Repository: openclaw/openclaw
• Author: zhanggpcsu
• State: closed | merged: False
• Link: https://github.com/openclaw/openclaw/pull/74327

Description (problem / solution / changelog)

Summary

Fixes #74205

Issue

[Bug]: kimi/k2p6 returns incomplete terminal response and gateway status probe times out

Solution

Testing

Changed files

• extensions/fireworks/index.test.ts (modified, +33/-0)
• extensions/fireworks/index.ts (modified, +9/-2)