Luis Pater 7c2f6ce0d1 fix(claude): avoid mid-conversation system splicing for advisor calls or results ...

- Detect advisor tool calls and advisor results in conversation history.
- Preserve forwarded system prompt blocks in top-level system instead of splicing them into message turns or prepending reminders.
- Prevent message index shifts that break layout bindings for encrypted advisor results and trigger upstream 400 errors.

Closes: #5470

2026-09-06 05:55:42 +08:00

..

helps

feat(claude): add Fable 5.1 reporting outcomes block and post-payload reconciliation

2026-09-04 15:23:31 +08:00

aistudio_executor_test.go

fix(aistudio): normalize thinking level to uppercase

2026-09-06 04:14:41 +08:00

aistudio_executor.go

fix(aistudio): normalize thinking level to uppercase

2026-09-06 04:14:41 +08:00

antigravity_executor_auth.go

fix(auth): preserve upstream status codes in antigravity auth errors

2026-09-01 08:37:02 +08:00

antigravity_executor_buildrequest_test.go

fix(auth): preserve upstream status codes in antigravity auth errors

2026-09-01 08:37:02 +08:00

antigravity_executor_credits_test.go

fix(antigravity): remove cross-endpoint fallback (#5209) (#5228)

2026-08-25 11:51:17 +08:00

antigravity_executor_credits.go

fix(antigravity): bypass quota cooldowns and credit hints when cooling is disabled

2026-09-02 22:01:55 +08:00

antigravity_executor_disable_cooling_test.go

fix(antigravity): bypass quota cooldowns and credit hints when cooling is disabled

2026-09-02 22:01:55 +08:00

antigravity_executor_execute.go

feat(auth): use home model capabilities for thinking

2026-09-03 13:10:30 +08:00

antigravity_executor_finish_reason_test.go

fix(antigravity): safely synthesize terminal finish reasons (#5230)

2026-08-25 15:13:04 +08:00

antigravity_executor_interactions_test.go

test(executor): update antigravity interactions test to use valid model

2026-09-01 18:00:04 +08:00

antigravity_executor_keepalive_test.go

fix(auth): prefer errors from upstream attempts

2026-08-29 12:50:46 +08:00

antigravity_executor_request.go

fix(antigravity): remove cross-endpoint fallback (#5209) (#5228)

2026-08-25 11:51:17 +08:00

antigravity_executor_signature_test.go

fix(antigravity): strip tool config, labels, and session id in token counting

2026-09-02 21:08:21 +08:00

antigravity_executor_split_usage_test.go

fix(antigravity): safely synthesize terminal finish reasons (#5230)

2026-08-25 15:13:04 +08:00

antigravity_executor_stream.go

feat(auth): use home model capabilities for thinking

2026-09-03 13:10:30 +08:00

antigravity_executor_tokens.go

feat(auth): use home model capabilities for thinking

2026-09-03 13:10:30 +08:00

antigravity_executor_transport_test.go

fix(antigravity): harden per-credential transport pooling

2026-08-11 18:33:50 +08:00

antigravity_executor.go

fix(claude): preserve tool pairing across intervening system messages

2026-08-31 19:27:20 +08:00

antigravity_home_model_capabilities_test.go

feat(auth): use home model capabilities for thinking

2026-09-03 13:10:30 +08:00

antigravity_preupstream_rewrite_differential_test.go

fix(gemini,antigravity): align parallel tool results with preceding tool calls

2026-08-25 14:57:46 +08:00

antigravity_preupstream_rewrite_legacy_oracle_test.go

fix(claude): preserve tool pairing across intervening system messages

2026-08-31 19:27:20 +08:00

antigravity_reasoning_replay_clear_test.go

fix(antigravity): preserve complete reasoning replay (#4525)

2026-07-25 14:07:31 +08:00

antigravity_reasoning_replay_index_test.go

perf(antigravity): batch reasoning replay mutations

2026-09-04 18:53:21 +08:00

antigravity_reasoning_replay_legacy_oracle_test.go

refactor(antigravity): route replay merge through the request index

2026-08-11 09:57:15 +08:00

antigravity_reasoning_replay_test.go

fix(antigravity): tighten replacement offset guard in tool provenance degradation

2026-09-04 09:45:49 +08:00

antigravity_reasoning_replay.go

perf(antigravity): batch reasoning replay mutations

2026-09-04 18:53:21 +08:00

antigravity_refresh_test.go

fix(auth): prefer errors from upstream attempts

2026-08-29 12:50:46 +08:00

antigravity_schema_sanitize_test.go

fix(gemini): strip 'encrypted' metadata from tool parameters schemas (#5065)

2026-08-19 19:13:56 +08:00

caching_verify_test.go

fix(claude): restore cloaked prompt-cache ownership and native shape

2026-08-11 19:33:40 +08:00

claude_executor_auth_race_test.go

fix(claude): replay measured OAuth wire, Fast and diagnostic profiles

2026-08-03 14:47:26 +08:00

claude_executor_auth_test.go

fix(claude): support setup-tokens and gracefully handle 403 OAuth profile errors (#4983) (#5083)

2026-08-20 00:15:31 +08:00

claude_executor_auth.go

fix(claude): support setup-tokens and gracefully handle 403 OAuth profile errors (#4983) (#5083)

2026-08-20 00:15:31 +08:00

claude_executor_beta_policy_test.go

feat(claude): support advisor tool beta header

2026-08-29 14:47:51 +08:00

claude_executor_cloaking.go

fix(claude): avoid mid-conversation system splicing for advisor calls or results

2026-09-06 05:55:42 +08:00

claude_executor_diagnostics_test.go

feat(claude): implement 2.1.258 billing header fingerprint chain and upstream request continuity

2026-09-04 15:23:31 +08:00

claude_executor_diagnostics.go

feat(claude): implement 2.1.258 billing header fingerprint chain and upstream request continuity

2026-09-04 15:23:31 +08:00

claude_executor_execute.go

feat(claude): harden probe and helper request classification, diagnostics isolation, and late cloaking

2026-09-04 15:23:31 +08:00

claude_executor_fable_ratelimit_test.go

fix(claude): treat allowed_warning as allowed in unified rate limit checks

2026-08-27 17:32:07 +08:00

claude_executor_fast_error_test.go

fix(claude): harden request lifecycle

2026-08-03 14:47:26 +08:00

claude_executor_fast_error.go

fix(claude): add Anthropic unified rate-limit parsing helpers

2026-08-15 14:08:16 +08:00

claude_executor_native_helper_test.go

fix(claude): upgrade default Claude Code baseline and fingerprint to 2.1.258

2026-09-03 10:21:13 +08:00

claude_executor_ratelimit_test.go

fix(auth): prefer errors from upstream attempts

2026-08-29 12:50:46 +08:00

claude_executor_request_bench_test.go

perf(claude): batch OAuth tool name rewrites

2026-08-03 14:47:26 +08:00

claude_executor_request_remap_test.go

fix(claude): restore tool_search_tool_result references and recognize advisor/agent server tools (#5044) (#5082)

2026-08-19 22:29:51 +08:00

claude_executor_request.go

feat(claude): add 2.1.258 dynamic beta headers, model fallbacks, and paired cache TTL

2026-09-04 15:23:31 +08:00

claude_executor_stream.go

feat(claude): harden probe and helper request classification, diagnostics isolation, and late cloaking

2026-09-04 15:23:31 +08:00

claude_executor_test.go

fix(claude): avoid mid-conversation system splicing for advisor calls or results

2026-09-06 05:55:42 +08:00

claude_executor_thinking_signature_test.go

test(claude): preserve adaptive thinking signatures

2026-08-03 14:47:26 +08:00

claude_executor_tokens.go

fix(claude): gate CCH signing by upstream origin and validate fingerprint-profile

2026-08-18 18:06:40 +08:00

claude_executor_wire_casing_test.go

fix(claude): send Anthropic header names with the real client's casing

2026-08-03 14:47:26 +08:00

claude_executor.go

fix(cliproxy): support base_URL-only config credentials and skip stale auth headers

2026-08-20 15:19:07 +08:00

claude_fingerprint_policy_test.go

fix(claude): gate CCH signing by upstream origin and validate fingerprint-profile

2026-08-18 18:06:40 +08:00

claude_fingerprint_policy.go

refactor(claude): deduplicate CLI identity application and tidy helpers

2026-08-18 18:21:33 +08:00

claude_mid_system_model_test.go

feat(claude): add fingerprint-profile=claude-code-cli for API keys and delegated providers (#5047)

2026-08-18 17:29:26 +08:00

claude_signing_test.go

fix(claude): gate CCH signing by upstream origin and validate fingerprint-profile

2026-08-18 18:06:40 +08:00

claude_signing.go

refactor(claude): deduplicate CLI identity application and tidy helpers

2026-08-18 18:21:33 +08:00

claude_thinking_replay_test.go

fix(claude): add Claude-compatible thinking replay persistence for multi-turn sessions

2026-08-09 04:36:39 +08:00

claude_thinking_replay.go

fix(claude): add Claude-compatible thinking replay persistence for multi-turn sessions

2026-08-09 04:36:39 +08:00

codex_executor_auth.go

chore(pluginhost): refactor and remove unused interceptors and executor methods

2026-07-26 14:31:45 +08:00

codex_executor_cache_test.go

fix(codex): normalize request session header to Session-Id and preload codex headers

2026-08-12 18:28:40 +08:00

codex_executor_compact_test.go

…

codex_executor_execute.go

feat(executor): support $-based custom headers from downstream request headers

2026-08-18 19:33:50 +08:00

codex_executor_grokbuild_keepalive_test.go

fix(codex): convert Grok client keepalive SSE frames to comments

2026-08-22 20:58:35 +08:00

codex_executor_imagegen_test.go

feat(executor): add normalization for parallel tool calls in Codex executors

2026-07-18 03:58:03 +08:00

codex_executor_input_ids_test.go

feature(codex): Normalize invalid sub2api message item IDs during Codex input sanitization

2026-08-03 02:45:49 +08:00

codex_executor_instructions_test.go

…

codex_executor_parallel_tool_calls_test.go

feat(executor): add normalization for parallel tool calls in Codex executors

2026-07-18 03:58:03 +08:00

codex_executor_reasoning_replay_cache_test.go

perf(codex): make replay prefix fingerprints incremental

2026-07-24 11:10:58 -07:00

codex_executor_reasoning.go

Merge PR #4546 into dev

2026-07-26 19:11:32 +08:00

codex_executor_request.go

fix(cliproxy): support base_URL-only config credentials and skip stale auth headers

2026-08-20 15:19:07 +08:00

codex_executor_retry_test.go

…

codex_executor_signature_test.go

…

codex_executor_spawn_agent_test.go

feat(codex): add model-level is-compat flag to rewrite MultiAgentV2 agent_message for Responses-compatible endpoints

2026-08-06 04:49:28 +08:00

codex_executor_stream_output_test.go

feat(codex): add opt-in stream bootstrap buffering and overload failover (#5115)

2026-08-20 21:31:53 +08:00

codex_executor_stream.go

feat(executor): measure effective TTFT with protocol-aware token classification (#5313)

2026-08-29 01:22:21 +08:00

codex_executor_terminal.go

feat(executor): measure effective TTFT with protocol-aware token classification (#5313)

2026-08-29 01:22:21 +08:00

codex_executor_tokens.go

feat(compat): preserve compat-mode thinking/signature blocks for API-key models

2026-08-06 17:19:24 +08:00

codex_executor_translate_test.go

…

codex_executor.go

chore(pluginhost): refactor and remove unused interceptors and executor methods

2026-07-26 14:31:45 +08:00

codex_openai_images_extract_test.go

…

codex_openai_images_test.go

feat(codex): add disable-codex-cloaking config option and refine header management

2026-07-30 16:37:40 +08:00

codex_openai_images.go

feat(executor): support $-based custom headers from downstream request headers

2026-08-18 19:33:50 +08:00

codex_stream_bootstrap_buffering_test.go

feat(codex): add opt-in stream bootstrap buffering and overload failover (#5115)

2026-08-20 21:31:53 +08:00

codex_websockets_connection.go

fix(auth): prefer errors from upstream attempts

2026-08-29 12:50:46 +08:00

codex_websockets_errors.go

chore(pluginhost): refactor and remove unused interceptors and executor methods

2026-07-26 14:31:45 +08:00

codex_websockets_execute.go

fix(auth): prefer errors from upstream attempts

2026-08-29 12:50:46 +08:00

codex_websockets_executor_store_test.go

feat: add Home credential concurrency support

2026-07-23 13:42:08 +08:00

codex_websockets_executor_test.go

fix(auth): prefer errors from upstream attempts

2026-08-29 12:50:46 +08:00

codex_websockets_executor.go

chore(pluginhost): refactor and remove unused interceptors and executor methods

2026-07-26 14:31:45 +08:00

codex_websockets_request.go

fix(cliproxy): support base_URL-only config credentials and skip stale auth headers

2026-08-20 15:19:07 +08:00

codex_websockets_session.go

fix(codex): clear multi-agent-v2 optimization state on namespace conflicts

2026-08-12 18:20:13 +08:00

codex_websockets_spawn_agent_test.go

fix(codex): clear multi-agent-v2 optimization state on namespace conflicts

2026-08-12 18:20:13 +08:00

codex_websockets_stream.go

fix(auth): prefer errors from upstream attempts

2026-08-29 12:50:46 +08:00

custom_magic_headers_test.go

feat(executor): support $-based custom headers from downstream request headers

2026-08-18 19:33:50 +08:00

executor_payload_optimization_test.go

perf(executor): replace sjson.SetBytes with optimized helpers for conditional payload updates

2026-07-20 13:57:14 +08:00

gemini_executor_signature_test.go

fix(auth): prefer errors from upstream attempts

2026-08-29 12:50:46 +08:00

gemini_executor_test.go

fix(auth): prefer errors from upstream attempts

2026-08-29 12:50:46 +08:00

gemini_executor.go

fix(auth): prefer errors from upstream attempts

2026-08-29 12:50:46 +08:00

gemini_vertex_executor.go

fix(auth): prefer errors from upstream attempts

2026-08-29 12:50:46 +08:00

home_codex_terminal_test.go

feat: add Home credential concurrency support

2026-07-23 13:42:08 +08:00

kimi_executor_test.go

fix(kimi): normalize tool and function parameter schemas

2026-08-29 01:22:22 +08:00

kimi_executor.go

refactor(kimi): use request thinking helper

2026-09-03 13:10:30 +08:00

kimi_home_model_capabilities_test.go

refactor(kimi): use request thinking helper

2026-09-03 13:10:30 +08:00

kimi_thinking_replay_test.go

fix(kimi): canonicalize K2.7 Code model aliases to official Kimi-For-Coding IDs

2026-08-08 05:08:42 +08:00

kimi_thinking_replay.go

fix(claude): add Claude-compatible thinking replay persistence for multi-turn sessions

2026-08-09 04:36:39 +08:00

openai_compat_executor_compact_test.go

fix(openai): handle premature SSE stream termination with terminal error emission

2026-08-12 13:05:06 +08:00

openai_compat_executor_reasoning_test.go

feat(compat): preserve Claude thinking/tool-call content for is-compat OpenAI compatibility models

2026-08-07 06:26:33 +08:00

openai_compat_executor_retry_test.go

fix(openai-compat): honor bounded rate-limit waits

2026-09-03 11:12:09 +08:00

openai_compat_executor_tool_results_test.go

feat(executor): normalize OpenAI tool results for text-only compatibility models

2026-08-03 03:37:24 +08:00

openai_compat_executor.go

fix(openai-compat): honor bounded rate-limit waits

2026-09-03 11:12:09 +08:00

openai_responses_signature_test.go

Reduce Codex request amplification for large payloads

2026-08-07 19:16:06 +08:00

openai_responses_signature.go

Reduce Codex request amplification for large payloads

2026-08-07 19:16:06 +08:00

websocket_lifecycle_bind_test.go

feat: add Home credential concurrency support

2026-07-23 13:42:08 +08:00

websocket_session_target_test.go

feat: add Home credential concurrency support

2026-07-23 13:42:08 +08:00

xai_executor_auth.go

chore(pluginhost): refactor and remove unused interceptors and executor methods

2026-07-26 14:31:45 +08:00

xai_executor_execute.go

fix(xai): fold namespace tools and restore dispatcher tool calls

2026-08-27 01:08:35 +08:00

xai_executor_media.go

fix(xai): use chat base url resolution for image and video requests

2026-08-30 14:01:18 +08:00

xai_executor_request.go

fix(xai): use chat base url resolution for image and video requests

2026-08-30 14:01:18 +08:00

xai_executor_response.go

fix(xai): fold namespace tools and restore dispatcher tool calls

2026-08-27 01:08:35 +08:00

xai_executor_stream.go

fix(xai): fold namespace tools and restore dispatcher tool calls

2026-08-27 01:08:35 +08:00

xai_executor_test.go

fix(xai): use chat base url resolution for image and video requests

2026-08-30 14:01:18 +08:00

xai_executor_tokens.go

chore(pluginhost): refactor and remove unused interceptors and executor methods

2026-07-26 14:31:45 +08:00

xai_executor.go

fix(xai): use chat base url resolution for image and video requests

2026-08-30 14:01:18 +08:00

xai_reasoning_replay.go

…

xai_status_err_test.go

fix(xai): remap bad-credentials 403s to unauthorized

2026-08-07 06:48:09 +08:00

xai_websockets_executor_test.go

fix(auth): prefer errors from upstream attempts

2026-08-29 12:50:46 +08:00

xai_websockets_executor.go

fix(auth): prefer errors from upstream attempts

2026-08-29 12:50:46 +08:00