v0.1.8

v0.1.8 Release Notes

Build and Test Agents

• Safety: Integrated NVIDIA as a safety provider.
• VectorDB: Added Qdrant as an inline provider.
• Agents: Added support for multiple tool groups in agents.
• Agents: Simplified imports for Agents in client package

Agent Evals and Model Customization

• Introduced DocVQA and IfEval benchmarks.

Deploying and Monitoring Agents

• Introduced a Containerfile and image workflow for the Playground.
• Implemented support for Bearer (API Key) authentication.
• Added attribute-based access control for resources.
• Fixes on docker deployments: use --pull always and standardized the default port to 8321
• Deprecated: /v1/inspect/providers use /v1/providers/ instead

Better Engineering

• Consolidated scripts under the ./scripts directory.
• Addressed mypy violations in various modules.
• Added Dependabot scans for Python dependencies.
• Implemented a scheduled workflow to update the changelog automatically.
• Enforced concurrency to reduce CI loads.

What's Changed

• docs: add additional guidance around using virtualenv by @nathan-weinberg in #1642
• Web updates to point to latest releases for Mobile SDK by @cmodi-meta in #1650
• docs: Fix trailing whitespace error by @kelbrown20 in #1669
• fix: passthrough impl response.content.text by @yyymeta in #1665
• fix: fixed import error by @cheesecake100201 in #1637
• fix: build distribution with podman by @jeffmaury in #1671
• docs: Fixing outputs in client cli and formatting suggestions by @kelbrown20 in #1668
• feat: added nvidia as safety provider by @cdgamarose-nv in #1248
• fix: add shutdown method for ProviderImpl by @nathan-weinberg in #1670
• chore: consolidate scripts under ./scripts directory by @booxter in #1646
• refactor: simplify command execution and remove PTY handling by @leseb in #1641
• ci: limit PR testing based on modified files by @nathan-weinberg in #1644
• feat(api): (1/n) datasets api clean up by @yanxi0830 in #1573
• docs: fixed broken tip in distro build docs by @nathan-weinberg in #1673
• chore: Remove unused cursor rules by @terrytangyuan in #1653
• ci: Add dependabot scans for Python deps by @terrytangyuan in #1618
• test: Bump slow_callback_duration to 200ms to avoid flaky remote vLLM unit tests by @terrytangyuan in #1675
• fix: agents with non-llama model by @ehhuang in #1550
• feat(agent): support multiple tool groups by @ehhuang in #1556
• fix: Add the option to not verify SSL at remote-vllm provider by @luis5tb in #1585
• ci(ollama): run more integration tests by @leseb in #1636
• feat: Created Playground Containerfile and Image Workflow by @jland-redhat in #1256
• feat: support nvidia hosted vision models (llama 3.2 11b/90b) by @mattf in #1278
• chore: enable ruff for ./scripts too by @booxter in #1643
• docs: fix broken test path in CONTRIBUTING.md by @nathan-weinberg in #1679
• feat: Qdrant inline provider by @dmartinol in #1273
• fix: return 4xx for non-existent resources in GET requests by @leseb in #1635
• chore: Add links and badges to both unit and integration tests by @terrytangyuan in #1632
• chore: Make code interpreter async by @cheesecake100201 in #1654
• ci: Add scheduled workflow to update changelog by @terrytangyuan in #1503
• docs: remove redundant installation instructions by @leseb in #1138
• chore: fix mypy violations in post_training modules by @booxter in #1548
• fix: Fixed import errors for UI and playground by @cheesecake100201 in #1666
• fix: avoid tensor memory error by @yyymeta in #1688
• feat(auth, rfc): Add support for Bearer (api_key) Authentication by @ashwinb in #1626
• docs: Remove mentions of focus on Llama models by @terrytangyuan in #1690
• chore: Add triagers list #1561 by @franciscojavierarceo in #1701
• docs: add documentation for RAGDocument by @ehhuang in #1693
• fix: Updating ToolCall.arguments to allow for json strings that can be decoded on client side by @hardikjshah in #1685
• fix: Call pandas.read_* in a seperate thread by @derekhiggins in #1698
• fix: fix open-benchmark template by @SLR722 in #1695
• fix: toolgroups unregister by @ehhuang in #1704
• feat: [new open benchmark] DocVQA by @yyymeta in #1647
• docs: update for mcp tools by @ehhuang in #1705
• fix: Correctly set CLI_ARGS using BUILD_PLATFORM env with llama stack… by @MichaelClifford in #1702
• feat: [New Eval Benchamark] IfEval by @SLR722 in #1708
• fix: telemetry logger by @cdoern in #1714
• chore: deprecate /v1/inspect/providers by @cdoern in #1678
• chore: Don't set type variables from register_schema() by @booxter in #1713
• fix: better tool call parsing error message by @ehhuang in #1710
• feat(server): add attribute based access control for resources by @ashwinb in #1703
• fix: update default tool call system prompt by @ehhuang in #1712
• fix: Restore discriminator for AlgorithmConfig by @booxter in #1706
• fix: Support types.UnionType in schemas by @booxter in #1721
• chore: mypy violations cleanup for inline::{telemetry,tool_runtime,vector_io} by @booxter in #1711
• feat(api): simplify client imports by @ehhuang in #1687
• fix: Add missing gcc in container build. Fixes #1716 by @terrytangyuan in #1727
• fix: tracing middleware to not start for lifespan events by @dineshyv in #1730
• chore: Don't hide stderr from api generator by @booxter in #1720
• docs: Add instruction on enabling tool calling for remote vLLM by @terrytangyuan in #1719
• fix: handle registry errors gracefully by @ashwinb in #1732
• fix: docker run with --pull always to fetch the latest image by @hardikjshah in #1733
• fix: Default to port 8321 everywhere by @hardikjshah in #1734
• feat: add different sinks for otel traces and metrics by @dineshyv in #1731
• fix: install pandas and numpy beforehand to avoid version mismatch by @hardikjshah in #1735
• ci: Enforce concurrency to reduce CI loads by @terrytangyuan in #1738
• fix: fix experimental-post-training template by @SLR722 in #1740
• fix: only invoke openapi generator if APIs or API generator changes by @ashwinb in #1744
• feat: make sure agent sessions are under access control by @ashwinb in #1737
• test: Add default vLLM URL in remote-vllm template by @terrytangyuan in #1736
• fix: Add 'accelerate' dependency to 'prompt-guard' by @derekhiggins in #1724
• docs: propose new contribution guidance by @leseb in #1750
• feat(telemetry): normalize path by @ehhuang in #1739
• docs: fix change dir command by @leseb in #1752
• docs: fix typos in evaluation concepts by @Bobbins228 in #1745
• feat(telemetry): use zero-width space to avoid clutter by @ehhuang in #1754
• fix: fix jobs api literal return type by @yanxi0830 in #1757
• chore(telemetry): remove service_name entirely by @ehhuang in #1755
• feat: use same trace ids in stack and otel by @dineshyv in #1759
• fix: Update getting_started.ipynb by @hardikjshah in #1761
• feat(telemetry): clean up spans by @ehhuang in #1760
• chore: make mypy happy with webmethod by @ehhuang in #1758

New Contributors

• @cmodi-meta made their first contribution in #1650
• @jeffmaury made their first contribution in #1671
• @derekhiggins made their first contribution in #1698
• @Bobbins228 made their first contribution in #1745

Full Changelog: v0.1.7...v0.1.8