Commit Graph

4763 Commits

Author SHA1 Message Date
Peter Steinberger
e9bf9fde06 test: split legacy pi-tools schema shards 2026-04-05 13:10:16 +01:00
Peter Steinberger
a060b89e3f fix(ci): remove duplicate grok test provider inference 2026-04-05 13:07:22 +01:00
Peter Steinberger
2635e07bf0 fix(openai): add multilingual gpt ack prompts 2026-04-05 12:57:41 +01:00
Peter Steinberger
76da484bed fix: infer synthetic provider auth in implicit tests 2026-04-05 12:54:09 +01:00
Peter Steinberger
deb212d3b0 fix(openai): tighten gpt chat action turns 2026-04-05 12:53:35 +01:00
Peter Steinberger
59a243e46b test: stabilize provider discovery matrix cases 2026-04-05 12:53:24 +01:00
Peter Steinberger
bc7f845714 test: speed up focused pi-tools tool tests 2026-04-05 12:53:15 +01:00
Peter Steinberger
dbcd35f6c2 test: decouple pi-tools params test imports 2026-04-05 12:53:15 +01:00
wirjo
78fe96f2d4 feat(bedrock): add inference profile discovery and region injection (#61299)
* feat(bedrock): add inference profile discovery and region injection

Inference profiles (cross-region and application) work with ConverseStream
but require the SDK client region to match the profile region. Without
this, users get "The provided model identifier is invalid" errors when
using cross-region profiles like us.anthropic.claude-sonnet-4-6.

Changes:

1. Inference profile discovery (discovery.ts):
   - Call ListInferenceProfiles alongside ListFoundationModels (parallel)
   - Inference profiles INHERIT capabilities from their underlying
     foundation model (modalities, reasoning, context window, cost)
   - resolveBaseModelId() maps profile → foundation model:
     "us.anthropic.claude-sonnet-4-6" → "anthropic.claude-sonnet-4-6"
     Application ARNs → extract model ID from models[].modelArn
   - Graceful degradation if IAM lacks bedrock:ListInferenceProfiles
   - Provider filter applies to profiles via underlying model ARNs

2. Region injection (register.sync.runtime.ts):
   - Extract region from provider baseUrl or bedrockDiscovery.region
   - Pass through to pi-ai options.region in wrapStreamFn
   - Ensures SDK client connects to correct regional endpoint

3. Inference profile model detection (anthropic-family-cache-semantics.ts):
   - isAnthropicBedrockModel() now recognizes application inference
     profile ARNs (arn:aws:bedrock:...:application-inference-profile/*)

4. Tests (discovery.test.ts):
   - New: inference profile inheritance test (4 models: 1 foundation +
     3 profiles, verifies capability inheritance, inactive filtering)
   - New: graceful AccessDeniedException handling test
   - Updated: all existing tests for dual-API discovery pattern

Fixes #55642

* fix(bedrock): preserve inference profile model lookup

---------

Co-authored-by: Vincent Koc <vincentkoc@ieee.org>
2026-04-05 12:52:03 +01:00
Peter Steinberger
996dccb19c feat(agents): add structured execution item events 2026-04-05 12:36:33 +01:00
Peter Steinberger
3b7e6152d1 fix: retry reasoning-required compaction with minimal thinking 2026-04-05 12:35:50 +01:00
Mariano
4175caee6e fix(agents): suppress commentary-phase output leaks (#61282)
Merged via squash.

Prepared head SHA: e392904f73
Co-authored-by: mbelinky <132747814+mbelinky@users.noreply.github.com>
Co-authored-by: mbelinky <132747814+mbelinky@users.noreply.github.com>
Reviewed-by: @mbelinky
2026-04-05 13:04:12 +02:00
Peter Steinberger
fef155cdbc fix: tighten file tool schemas for openai 2026-04-05 11:46:34 +01:00
Peter Steinberger
064f474ed7 test: isolate anthropic vertex provider env cases 2026-04-05 11:28:59 +01:00
Peter Steinberger
4559ece355 fix(ci): align test fixtures with current runner types 2026-04-05 11:23:51 +01:00
Chunyue Wang
8c1ca1f245 fix(cron): remove OpenAPI 3.0 incompatible JSON Schema keywords from cron tool (#61221)
The cron tool schema used type arrays (['string','null']), the 'not'
keyword, and 'const' — all unsupported by the OpenAPI 3.0 subset that
Gemini-backed providers (e.g. GitHub Copilot) enforce. This caused
HTTP 400 for every request when cron was enabled.

Replace type arrays with scalar types, remove not/const from
CronFailureAlertSchema, and add 'not' to the Gemini unsupported
keywords list as defense-in-depth.

Fixes #61206
2026-04-05 11:21:45 +01:00
Peter Steinberger
af81ee9fee fix(agents): add embedded item lifecycle events 2026-04-05 11:16:28 +01:00
Peter Steinberger
fd917a471c test: scope implicit provider discovery harness 2026-04-05 11:12:41 +01:00
Vincent Koc
3b84884793 fix(agents): harden host-managed claude-cli auth path (#61276) 2026-04-05 11:02:18 +01:00
Peter Steinberger
afca9540bf fix: add openai responses phase support 2026-04-05 10:58:49 +01:00
Vincent Koc
1d736dcbbc fix(ci): drop unused google prompt cache type 2026-04-05 10:49:51 +01:00
Vincent Koc
64cf52ca20 fix(tool-display): generate swift snapshot from core config 2026-04-05 10:34:02 +01:00
Peter Steinberger
e468da1040 fix: improve gpt execution flow and visibility 2026-04-05 10:32:58 +01:00
Tuyen
cd348659ce Anthropic: fix claude-cli runtime auth 2026-04-05 18:19:33 +09:00
Peter Steinberger
9d315cdf42 test: default vitest lanes to isolated forks 2026-04-05 10:11:33 +01:00
Vincent Koc
d5cde2171b fix(agents): surface disk full session write errors (#61264) 2026-04-05 10:09:42 +01:00
Vincent Koc
cc09171929 fix(ci): align sanitize session history tests with transcript types 2026-04-05 10:01:29 +01:00
Vincent Koc
4613f121ad fix(agents): preserve native Anthropic replay tool ids (#61254)
* fix(agents): preserve native Anthropic replay tool ids

* docs(changelog): note native Anthropic replay ids

* fix(agents): preserve native Anthropic replay ids selectively
2026-04-05 09:53:52 +01:00
Peter Steinberger
3d952aa35d fix(agents): preserve claude cli backend defaults 2026-04-05 17:51:41 +09:00
Peter Steinberger
48653c2031 fix: recover launchd restart and restore prompt-cache gate 2026-04-05 17:47:07 +09:00
Peter Steinberger
48f0d9aaf7 fix: normalize telegram announce thread targets 2026-04-05 09:44:20 +01:00
Peter Steinberger
22db77d2b6 fix: avoid eager web provider config reads 2026-04-05 09:34:01 +01:00
Vincent Koc
2d6e75ccd5 fix(ci): align google prompt cache stream typing 2026-04-05 09:31:39 +01:00
Peter Steinberger
bff55b55cb style: normalize import ordering and wrapping 2026-04-05 09:26:39 +01:00
rstar327
43fe68f9ef fix(exec): keep notifyOnExit heartbeat wakes on exec-event (#41479)
Co-authored-by: Vincent Koc <vincentkoc@ieee.org>
2026-04-05 09:25:12 +01:00
Peter Steinberger
8be017fae6 refactor: remove plugin sdk facade generator 2026-04-05 09:23:55 +01:00
Peter Steinberger
5ac07b8ef0 fix: normalize huggingface refs and discovery timeout 2026-04-05 09:18:17 +01:00
Vincent Koc
b5f8cd4fcf fix(google): add managed gemini prompt caching 2026-04-05 09:17:51 +01:00
Peter Steinberger
3038079c2f fix: clamp pi embedded fenced chunk splits 2026-04-05 09:09:26 +01:00
Peter Steinberger
1903be5401 refactor: remove generated plugin sdk facades 2026-04-05 09:07:33 +01:00
Peter Steinberger
07e7b7177f test: stabilize pi embedded text-end hooks 2026-04-05 09:06:15 +01:00
Peter Steinberger
e28e83f4e4 test: load browser fixture for sandbox policy 2026-04-05 09:00:42 +01:00
Peter Steinberger
3728cfbe29 test: use stable sandbox denied tool assertions 2026-04-05 08:58:55 +01:00
Peter Steinberger
70b8ce72df test: refresh simple completion provider fallback 2026-04-05 08:57:04 +01:00
Peter Steinberger
23275edef1 refactor: simplify web provider plugin discovery 2026-04-05 08:50:01 +01:00
Vincent Koc
87b8680ded fix(cache): order stable project context before heartbeat (#61236)
* fix(cache): order stable project context before heartbeat

* docs(changelog): note project context cache ordering

* Update CHANGELOG.md
2026-04-05 08:49:20 +01:00
Peter Steinberger
07c2f81392 fix: preserve explicit Ollama apiKey during discovery 2026-04-05 08:43:50 +01:00
Peter Steinberger
acd8966ff0 test: refresh agent model expectation fixtures 2026-04-05 08:33:54 +01:00
Peter Steinberger
31d8b022eb fix: treat inline buttons as native approval ui 2026-04-05 08:33:54 +01:00
Ted Li
b474e098d1 docs: correct overstated prompt-cache comments from #58036 #58037 #58038 (#60633)
* docs: correct overstated prompt-cache comments from #58036 #58037 #58038

* docs: restore purpose context in MCP tool sort comment

* docs: drop misleading 'legacy' framing from image-prune comments

* docs: restore useful context stripped from image-prune comments

* docs: restore 'deterministically' in MCP tool sort comment

* docs: restore 'idempotent' at attempt.ts callsite

* docs: restore 'provider prompt cache' in context-guard comment
2026-04-05 08:32:51 +01:00