v0.1.5

0.1.5 Release Notes

Build Agents

• Inference: Support more non-llama models (openai, anthropic, gemini)
• Inference: Can use the provider's model name in addition to the HF alias
• Inference: Fixed issues with calling tools that weren't specified in the prompt
• RAG: Improved system prompt for RAG and no more need for hard-coded rag-tool calling
• Embeddings: Added support for Nemo retriever embedding models
• Tools: Added support for MCP tools in Ollama Distribution
• Distributions: Added new Groq distribution

Customize Models

• Save post-trained checkpoint in SafeTensor format to allow Ollama inference provider to use the post-trained model

Monitor agents

• More comprehensive logging of agent steps including client tools
• Telemetry inputs/outputs are now structured and queryable
• Ability to retrieve agents session, turn, step by ids

Better Engineering

• Moved executorch Swift code out of this repo into the llama-stack-client-swift repo, similar to kotlin
• Move most logging to use logger instead of prints
• Completed text /chat-completion and /completion tests

All changes

• test: add a ci-tests distro template for running e2e tests by @ashwinb in #1237
• refactor: combine start scripts for each env by @cdoern in #1139
• fix: pre-commit updates by @cdoern in #1243
• fix: Update getting_started.ipynb by @hardikjshah in #1245
• fix: Update Llama_Stack_Benchmark_Evals.ipynb by @hardikjshah in #1246
• build: hint on Python version for uv venv by @leseb in #1172
• fix: include timezone in Agent steps' timestamps by @ehhuang in #1247
• LocalInferenceImpl update for LS013 by @jeffxtang in #1242
• fix: Raise exception when tool call result is None by @terrytangyuan in #1253
• fix: resolve type hint issues and import dependencies by @leseb in #1176
• fix: build_venv expects an extra argument by @cdoern in #1233
• feat: completing text /chat-completion and /completion tests by @LESSuseLESS in #1223
• fix: update index.md to include 0.1.4 by @raghotham in #1259
• docs: Remove $ from client CLI ref to add valid copy and paste ability by @kelbrown20 in #1260
• feat: Add Groq distribution template by @VladOS95-cyber in #1173
• chore: update the zero_to_hero_guide doc link by @reidliu41 in #1220
• build: Merge redundant "files" field for codegen check in .pre-commit-config.yaml by @terrytangyuan in #1261
• refactor(server): replace print statements with logger by @leseb in #1250
• fix: fix the describe table display issue by @reidliu41 in #1221
• chore: update download error message by @reidliu41 in #1217
• chore: removed executorch submodule by @jeffxtang in #1265
• refactor: move OpenAI compat utilities from nvidia to openai_compat by @ashwinb in #1258
• feat: add (openai, anthropic, gemini) providers via litellm by @ashwinb in #1267
• feat: [post training] support save hf safetensor format checkpoint by @SLR722 in #845
• fix: the pre-commit new line issue by @reidliu41 in #1272
• fix(cli): Missing default for --image-type in stack run command by @terrytangyuan in #1274
• fix: Get builtin tool calling working in remote-vllm by @bbrowning in #1236
• feat: remove special handling of builtin::rag tool by @ehhuang in #1015
• feat: update the post training notebook by @SLR722 in #1280
• fix: time logging format by @ehhuang in #1281
• feat: allow specifying specific tool within toolgroup by @ehhuang in #1239
• fix: sqlite conn by @ehhuang in #1282
• chore: upgrade uv pre-commit version, uv-sync -> uv-lock by @ashwinb in #1284
• fix: don't attempt to clean gpu memory up when device is cpu by @booxter in #1191
• feat: Add model context protocol tools with ollama provider by @Shreyanand in #1283
• fix(test): update client-sdk tests to handle tool format parametrization better by @ashwinb in #1287
• feat: add nemo retriever text embedding models to nvidia inference provider by @mattf in #1218
• feat: don't silently ignore incorrect toolgroup by @ehhuang in #1285
• feat: ability to retrieve agents session, turn, step by ids by @yanxi0830 in #1286
• fix(test): no need to specify tool prompt format explicitly in tests by @ashwinb in #1295
• chore: remove vector_db_id from AgentSessionInfo by @yanxi0830 in #1296
• fix: Revert "chore: remove vector_db_id from AgentSessionInfo" by @yanxi0830 in #1299
• feat(providers): Groq now uses LiteLLM openai-compat by @ashwinb in #1303
• fix: duplicate ToolResponseMessage in Turn message history by @yanxi0830 in #1305
• fix: don't include tool args not in the function definition by @ehhuang in #1307
• fix: update notebooks to avoid using the nutsy --image-name system thing by @ashwinb in #1308
• fix: register provider model name and HF alias in run.yaml by @ashwinb in #1304
• build: Add dotenv file for running tests with uv by @terrytangyuan in #1251
• docs: update the output of llama-stack-client models list by @reidliu41 in #1271
• fix: Avoid unexpected keyword argument for sentence_transformers by @luis5tb in #1269
• feat: add nvidia embedding implementation for new signature, task_type, output_dimention, text_truncation by @mattf in #1213
• chore: add subcommands description in help by @reidliu41 in #1219
• fix: Structured outputs for recursive models by @hardikjshah in #1311
• fix: litellm tool call parsing event type to in_progress by @yanxi0830 in #1312
• fix: Incorrect import path for print_subcommand_description() by @terrytangyuan in #1313
• fix: Incorrect import path for print_subcommand_description() by @terrytangyuan in #1314
• fix: Incorrect import path for print_subcommand_description() by @terrytangyuan in #1315
• test: Only run embedding tests for remote::nvidia by @terrytangyuan in #1317
• fix: update getting_started notebook to pass nbeval by @yanxi0830 in #1318
• fix: [Litellm]Do not swallow first token by @hardikjshah in #1316
• feat: update the default system prompt for 3.2/3.3 models by @ehhuang in #1310
• fix: Agent telemetry inputs/outputs should be structured by @hardikjshah in #1302
• fix: check conda env name using basepath in exec.py by @dineshyv in #1301

New Contributors

• @Shreyanand made their first contribution in #1283
• @luis5tb made their first contribution in #1269

Full Changelog: v0.1.4...v0.1.5