[WRONG BRANCH] fix: stabilize cache inputs - #625

Closed

jgautheron wants to merge 33 commits into

lidge-jun:mainfrom

jgautheron:fix/gpt-codex-cache-stability-pr

Commits on Jul 27, 2026

fix(openai-responses): repair oversized replay call_ids on non-forward routes too
jgautheron
committed
3a0b88f
fix(claude/inbound): prefer content-based prompt_cache_key over session-scoped
jgautheron
committed
41deeef
fix(cli): ocx restart no longer silently no-ops when codexAutoStart is disabled
jgautheron
committed
23ec63f
Merge fix/call-id-oversized-replay into main
jgautheron
committed
ad58ac8
Merge fix/cross-session-cache-key-priority into main
jgautheron
committed
0215404
Merge fix/restart-ignores-codex-autostart into main
jgautheron
committed
44d0294
fix: serialize parallel subagent dispatches sharing a cold prompt_cache_key
jgautheron
committed
152682b
Merge fix/cold-cache-key-parallel-lease into main
jgautheron
committed
627e43c
feat(router): warn when a provider's bare gpt-/o1-/o3-/o4- model is unreachable
jgautheron
committed
5823712
Merge fix/warn-unreachable-bare-model-aliases into main
jgautheron
committed
4430ed3
Merge remote-tracking branch 'upstream/dev'
jgautheron
committed
d9a2a55
feat(router): extend bare-model warning to MODEL_PROVIDER_PATTERNS collisions
jgautheron
committed
9ce0464
Merge fix/warn-unreachable-bare-model-aliases (pattern-table extension) into main
jgautheron
committed
00b3861
fix(cli): restart/tray-start lost the Grok fence + Codex model sync
jgautheron
committed
ff0596d
test(claude-inbound): fix devlog 130 B3 test to match content-first cache-key priority
jgautheron
committed
e20c21e
Merge branch 'fix/cache-key-provenance-test-parity'
jgautheron
committed
8298112
docs(router): simplify bare-model-alias warning comment
jgautheron
committed
b001af9
docs: simplify cold-cache-key-lease comments
jgautheron
committed
7b0fb67
docs(claude/inbound): simplify cache-key-priority comments
jgautheron
committed
ba2085b
docs: simplify restart/grok-sync-parity comments
jgautheron
committed
2f3a7dd
docs(claude-inbound): simplify cache-key test comments
jgautheron
committed
b4f95ae
Merge branch 'fix/warn-unreachable-bare-model-aliases'
jgautheron
committed
bae00dd
Merge branch 'fix/cold-cache-key-parallel-lease'
jgautheron
committed
4da468a
Merge branch 'fix/cross-session-cache-key-priority'
jgautheron
committed
479390e
Merge branch 'fix/restart-grok-model-sync-parity'
jgautheron
committed
a0972c2
Merge branch 'fix/cache-key-provenance-test-parity'
jgautheron
committed
11cdee2
fix(usage): generic wildcard overlay fallback for private-gateway pricing
jgautheron
committed
d637087
fix(adapters): extract cached_tokens from Responses API usage
jgautheron
committed
3af6d4a
test(router): rename ambiguous fixture provider key
jgautheron
committed
bfda29d
Merge branch 'fix/openai-responses-cached-tokens'
jgautheron
committed
d99a373
Merge branch 'fix/warn-test-fixture-provider-naming'
jgautheron
committed
b8ddda0

Commits on Jul 28, 2026

fix(claude): stabilize routed cache inputs
jgautheron
committed
8cc4747
fix(claude): preserve native cache session affinity
jgautheron
committed
9e89c41

Navigation Menu

[WRONG BRANCH] fix: stabilize cache inputs - #625

[WRONG BRANCH] fix: stabilize cache inputs#625
jgautheron wants to merge 33 commits into
lidge-jun:mainfrom
jgautheron:fix/gpt-codex-cache-stability-pr

Commits on Jul 27, 2026

fix(openai-responses): repair oversized replay call_ids on non-forward routes too

fix(claude/inbound): prefer content-based prompt_cache_key over session-scoped

fix(cli): ocx restart no longer silently no-ops when codexAutoStart is disabled

Merge fix/call-id-oversized-replay into main

Merge fix/cross-session-cache-key-priority into main

Merge fix/restart-ignores-codex-autostart into main

fix: serialize parallel subagent dispatches sharing a cold prompt_cache_key

Merge fix/cold-cache-key-parallel-lease into main

feat(router): warn when a provider's bare gpt-/o1-/o3-/o4- model is unreachable

Merge fix/warn-unreachable-bare-model-aliases into main

Merge remote-tracking branch 'upstream/dev'

feat(router): extend bare-model warning to MODEL_PROVIDER_PATTERNS collisions

Merge fix/warn-unreachable-bare-model-aliases (pattern-table extension) into main

fix(cli): restart/tray-start lost the Grok fence + Codex model sync

test(claude-inbound): fix devlog 130 B3 test to match content-first cache-key priority

Merge branch 'fix/cache-key-provenance-test-parity'

docs(router): simplify bare-model-alias warning comment

docs: simplify cold-cache-key-lease comments

docs(claude/inbound): simplify cache-key-priority comments

docs: simplify restart/grok-sync-parity comments

docs(claude-inbound): simplify cache-key test comments

Merge branch 'fix/warn-unreachable-bare-model-aliases'

Merge branch 'fix/cold-cache-key-parallel-lease'

Merge branch 'fix/cross-session-cache-key-priority'

Merge branch 'fix/restart-grok-model-sync-parity'

Merge branch 'fix/cache-key-provenance-test-parity'

fix(usage): generic wildcard overlay fallback for private-gateway pricing

fix(adapters): extract cached_tokens from Responses API usage

test(router): rename ambiguous fixture provider key

Merge branch 'fix/openai-responses-cached-tokens'

Merge branch 'fix/warn-test-fixture-provider-naming'

Commits on Jul 28, 2026

fix(claude): stabilize routed cache inputs

fix(claude): preserve native cache session affinity