| 文件 | 最后提交记录 | 最后更新时间 |
|---|---|---|
fix(server): report stage-router stats for a composite route (#943) AlgorithmStats::new creates the stage-router baseline only when a route's algorithm name is literally "stage_router". A composite route is a stage route under another name: CompositeRouter is build_stage_route(config.stage)?.with_name(COMPOSITE) and its StageClassifier records switchyard.stage_router.routing_decisions and the scoring histograms on every turn regardless of the route name. With no baseline the snapshot field stays None and serde skips it, so an operator running a composite route gets algorithm_stats: {} from /v1/stats while /metrics carries the counters, and loses the decision-source breakdown the stage-router docs promise. Key the baseline on either name. The wider fix is for libsy to say which algorithms a route composes, so the next wrapper does not reintroduce this; this keeps the change inside the server, next to the two names already hardcoded there. Signed-off-by: Dmitry Voropaev <dy.voropaev@gmail.com> | 6 天前 | |
fix(server): reject inputs disabled by route capabilities (#750) * fix(server): reject inputs disabled by route capabilities Signed-off-by: Elyas Mehtabuddin <emehtabuddin@nvidia.com> * fix(server): reject MCP approval items when tool calling is disabled Signed-off-by: Elyas Mehtabuddin <emehtabuddin@nvidia.com> * chore(server): simplify capability tests Signed-off-by: Elyas Mehtabuddin <emehtabuddin@nvidia.com> --------- Signed-off-by: Elyas Mehtabuddin <emehtabuddin@nvidia.com> | 1 个月前 | |
feat(costs): use bundled model prices by default (#1049) Signed-off-by: Elyas Mehtabuddin <emehtabuddin@nvidia.com> | 13 小时前 | |
feat: Extract from switchyard-server into switchyard-runner (#517) * feat: Extract from switchyard-server into switchyard-runner These pieces will be shared by integrations, particularly the NeMo-Relay plugin. Implements: https://github.com/NVIDIA-NeMo/Switchyard/issues/514 Assisted-by: Codex:5.6 Sol high Assisted-by: Claude:Opus 5 medium Signed-off-by: Graham King <grahamk@nvidia.com> | 1 个月前 | |
feat(costs): save call costs and share pricing (#1045) * feat(server): record each call's cost in the routing log Signed-off-by: Elyas Mehtabuddin <emehtabuddin@nvidia.com> * chore(costs): simplify pricing and consolidate coverage Signed-off-by: Elyas Mehtabuddin <emehtabuddin@nvidia.com> * chore(server): keep pricing private and simplify session totals Signed-off-by: Elyas Mehtabuddin <emehtabuddin@nvidia.com> * chore(server): move Anthropic cache pricing into its own module Signed-off-by: Elyas Mehtabuddin <emehtabuddin@nvidia.com> * chore(costs): remove added tests and the standalone guide Signed-off-by: Elyas Mehtabuddin <emehtabuddin@nvidia.com> * chore(docs): restore pricing-file setup instructions Signed-off-by: Elyas Mehtabuddin <emehtabuddin@nvidia.com> * fix(costs): retain routing charges and preserve Chat usage fields Signed-off-by: Elyas Mehtabuddin <emehtabuddin@nvidia.com> * feat(costs): share recorded spending with the menubar Signed-off-by: Elyas Mehtabuddin <emehtabuddin@nvidia.com> --------- Signed-off-by: Elyas Mehtabuddin <emehtabuddin@nvidia.com> | 1 天前 | |
feat(server): add native Python host with OpenTelemetry (#210) * feat(server): add native Python host with OpenTelemetry Signed-off-by: nachiketb <nachiketb@nvidia.com> * fix(server): validate embedded host configuration Signed-off-by: nachiketb <nachiketb@nvidia.com> --------- Signed-off-by: nachiketb <nachiketb@nvidia.com> | 2 个月前 | |
feat(observability): expose translation diagnostics (#608) * feat(observability): expose translation diagnostics Signed-off-by: Clement Pakkam Isaac <cpakkamisaac@nvidia.com> * test(observability): tighten translation diagnostic coverage Signed-off-by: Clement Pakkam Isaac <cpakkamisaac@nvidia.com> * refactor(observability): keep translation diagnostics internal Signed-off-by: Clement Pakkam Isaac <cpakkamisaac@nvidia.com> * refactor(observability): distinguish observed translation wrappers Signed-off-by: Clement Pakkam Isaac <cpakkamisaac@nvidia.com> * refactor(observability): share translation diagnostic recorder Signed-off-by: Clement Pakkam Isaac <cpakkamisaac@nvidia.com> --------- Signed-off-by: Clement Pakkam Isaac <cpakkamisaac@nvidia.com> | 6 天前 | |
test(server): fix unit tests that sometimes capture no log events (#895) Signed-off-by: Elyas Mehtabuddin <emehtabuddin@nvidia.com> | 16 天前 | |
feat(observability): expose translation diagnostics (#608) * feat(observability): expose translation diagnostics Signed-off-by: Clement Pakkam Isaac <cpakkamisaac@nvidia.com> * test(observability): tighten translation diagnostic coverage Signed-off-by: Clement Pakkam Isaac <cpakkamisaac@nvidia.com> * refactor(observability): keep translation diagnostics internal Signed-off-by: Clement Pakkam Isaac <cpakkamisaac@nvidia.com> * refactor(observability): distinguish observed translation wrappers Signed-off-by: Clement Pakkam Isaac <cpakkamisaac@nvidia.com> * refactor(observability): share translation diagnostic recorder Signed-off-by: Clement Pakkam Isaac <cpakkamisaac@nvidia.com> --------- Signed-off-by: Clement Pakkam Isaac <cpakkamisaac@nvidia.com> | 6 天前 | |
fix(relay): redact configured provider keys from plugin outputs (#738) * fix(relay): redact configured provider keys from plugin outputs Signed-off-by: Bryan Bednarski <bbednarski@nvidia.com> * fix(redaction): match JSON-escaped credentials in string values Signed-off-by: Bryan Bednarski <bbednarski@nvidia.com> --------- Signed-off-by: Bryan Bednarski <bbednarski@nvidia.com> | 1 个月前 | |
feat(observability): expose translation diagnostics (#608) * feat(observability): expose translation diagnostics Signed-off-by: Clement Pakkam Isaac <cpakkamisaac@nvidia.com> * test(observability): tighten translation diagnostic coverage Signed-off-by: Clement Pakkam Isaac <cpakkamisaac@nvidia.com> * refactor(observability): keep translation diagnostics internal Signed-off-by: Clement Pakkam Isaac <cpakkamisaac@nvidia.com> * refactor(observability): distinguish observed translation wrappers Signed-off-by: Clement Pakkam Isaac <cpakkamisaac@nvidia.com> * refactor(observability): share translation diagnostic recorder Signed-off-by: Clement Pakkam Isaac <cpakkamisaac@nvidia.com> --------- Signed-off-by: Clement Pakkam Isaac <cpakkamisaac@nvidia.com> | 6 天前 | |
feat(costs): use bundled model prices by default (#1049) Signed-off-by: Elyas Mehtabuddin <emehtabuddin@nvidia.com> | 13 小时前 | |
fix(server): handle graceful process shutdown (#294) * fix(server): handle graceful process shutdown Signed-off-by: nachiketb <nachiketb@nvidia.com> * refactor(server): isolate platform shutdown signals Signed-off-by: nachiketb <nachiketb@nvidia.com> * refactor(server): await shutdown without spawning Signed-off-by: nachiketb <nachiketb@nvidia.com> --------- Signed-off-by: nachiketb <nachiketb@nvidia.com> | 2 个月前 | |
fix(translation): finish Codex tool calls through a LiteLLM gateway (#969) * fix(translation): finish Codex tool calls and compaction through a LiteLLM gateway Signed-off-by: Elyas Mehtabuddin <emehtabuddin@nvidia.com> * test(server): shrink the LiteLLM gateway regression tests Signed-off-by: Elyas Mehtabuddin <emehtabuddin@nvidia.com> * refactor: simplify replayed tool definitions, custom-call decoding, and gateway tests Signed-off-by: Elyas Mehtabuddin <emehtabuddin@nvidia.com> * test: move the snapshot-conflict check to the translation tests Signed-off-by: Elyas Mehtabuddin <emehtabuddin@nvidia.com> * fix: stop adding tool definitions to requests that replay tool calls Signed-off-by: Elyas Mehtabuddin <emehtabuddin@nvidia.com> --------- Signed-off-by: Elyas Mehtabuddin <emehtabuddin@nvidia.com> | 6 天前 | |
feat(server): expose stage router stats (#344) * feat(server): expose stage router stats Signed-off-by: nachiketb <nachiketb@nvidia.com> * refactor(server): consolidate stats ownership Signed-off-by: nachiketb <nachiketb@nvidia.com> * docs: preserve released known issues Signed-off-by: nachiketb <nachiketb@nvidia.com> * test(server): colocate stage router stats coverage Signed-off-by: nachiketb <nachiketb@nvidia.com> * refactor(server): simplify algorithm stats state Signed-off-by: nachiketb <nachiketb@nvidia.com> * refactor(server): simplify stage router stats state Signed-off-by: nachiketb <nachiketb@nvidia.com> * refactor(server): isolate algorithm stats extensions Signed-off-by: nachiketb <nachiketb@nvidia.com> * refactor(metrics): colocate stage histogram boundaries Signed-off-by: nachiketb <nachiketb@nvidia.com> * refactor(stats): simplify stage metric summaries Signed-off-by: nachiketb <nachiketb@nvidia.com> --------- Signed-off-by: nachiketb <nachiketb@nvidia.com> | 2 个月前 | |
test(server): fix unit tests that sometimes capture no log events (#895) Signed-off-by: Elyas Mehtabuddin <emehtabuddin@nvidia.com> | 16 天前 | |
feat(costs): save call costs and share pricing (#1045) * feat(server): record each call's cost in the routing log Signed-off-by: Elyas Mehtabuddin <emehtabuddin@nvidia.com> * chore(costs): simplify pricing and consolidate coverage Signed-off-by: Elyas Mehtabuddin <emehtabuddin@nvidia.com> * chore(server): keep pricing private and simplify session totals Signed-off-by: Elyas Mehtabuddin <emehtabuddin@nvidia.com> * chore(server): move Anthropic cache pricing into its own module Signed-off-by: Elyas Mehtabuddin <emehtabuddin@nvidia.com> * chore(costs): remove added tests and the standalone guide Signed-off-by: Elyas Mehtabuddin <emehtabuddin@nvidia.com> * chore(docs): restore pricing-file setup instructions Signed-off-by: Elyas Mehtabuddin <emehtabuddin@nvidia.com> * fix(costs): retain routing charges and preserve Chat usage fields Signed-off-by: Elyas Mehtabuddin <emehtabuddin@nvidia.com> * feat(costs): share recorded spending with the menubar Signed-off-by: Elyas Mehtabuddin <emehtabuddin@nvidia.com> --------- Signed-off-by: Elyas Mehtabuddin <emehtabuddin@nvidia.com> | 1 天前 |
| 文件 | 最后提交记录 | 最后更新时间 |
|---|---|---|
| 6 天前 | ||
| 1 个月前 | ||
| 13 小时前 | ||
| 1 个月前 | ||
| 1 天前 | ||
| 2 个月前 | ||
| 6 天前 | ||
| 16 天前 | ||
| 6 天前 | ||
| 1 个月前 | ||
| 6 天前 | ||
| 13 小时前 | ||
| 2 个月前 | ||
| 6 天前 | ||
| 2 个月前 | ||
| 16 天前 | ||
| 1 天前 |