phoenix-oss/llama-stack
1
0
Fork 0
forked from phoenix-oss/llama-stack-mirror
Code Pull requests Releases Packages 2 Activity Actions

feat: add a configurable category-based logger (#1352)

Browse source 
A self-respecting server needs good observability which starts with
configurable logging. Llama Stack had little until now. This PR adds a
`logcat` facility towards that. Callsites look like:

```python
logcat.debug("inference", f"params to ollama: {params}")
```

- the first parameter is a category. there is a static list of
categories in `llama_stack/logcat.py`
- each category can be associated with a log-level which can be
configured via the `LLAMA_STACK_LOGGING` env var.
- a value `LLAMA_STACK_LOGGING=inference=debug;server=info"` does the
obvious thing. there is a special key called `all` which is an alias for
all categories

## Test Plan

Ran with `LLAMA_STACK_LOGGING="all=debug" llama stack run fireworks` and
saw the following:


![image](https://github.com/user-attachments/assets/d24b95ab-3941-426c-9ea0-a4c62542e6f0)

Hit it with a client-sdk test case and saw this:


![image](https://github.com/user-attachments/assets/3fee8c6c-986e-4125-a09c-f5dc019682e2)
This commit is contained in:
Ashwin Bharambe 2025-03-02 18:51:14 -08:00 • committed by GitHub
parent a9a7b11326
commit 754feba61f
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
12 changed files with 409 additions and 47 deletions
Download patch file Download diff file Expand all files Collapse all files

60
llama_stack/distribution/routers/routers.py
View file

@ -7,6 +7,7 @@
import copy
from typing import Any, AsyncGenerator, Dict, List, Optional
from llama_stack import logcat
from llama_stack.apis.common.content_types import (
URL,
InterleavedContent,
@ -63,12 +64,15 @@ class VectorIORouter(VectorIO):
self,
routing_table: RoutingTable,
) -> None:
logcat.debug("core", "Initializing VectorIORouter")
self.routing_table = routing_table
async def initialize(self) -> None:
logcat.debug("core", "VectorIORouter.initialize")
pass
async def shutdown(self) -> None:
logcat.debug("core", "VectorIORouter.shutdown")
pass
async def register_vector_db(
@ -79,6 +83,7 @@ class VectorIORouter(VectorIO):
provider_id: Optional[str] = None,
provider_vector_db_id: Optional[str] = None,
) -> None:
logcat.debug("core", f"VectorIORouter.register_vector_db: {vector_db_id}, {embedding_model}")
await self.routing_table.register_vector_db(
vector_db_id,
embedding_model,
@ -93,6 +98,10 @@ class VectorIORouter(VectorIO):
chunks: List[Chunk],
ttl_seconds: Optional[int] = None,
) -> None:
logcat.debug(
"core",
f"VectorIORouter.insert_chunks: {vector_db_id}, {len(chunks)} chunks, ttl_seconds={ttl_seconds}, chunk_ids={[chunk.id for chunk in chunks[:3]]}{' and more...' if len(chunks) > 3 else ''}",
)
return await self.routing_table.get_provider_impl(vector_db_id).insert_chunks(vector_db_id, chunks, ttl_seconds)
async def query_chunks(
@ -101,6 +110,7 @@ class VectorIORouter(VectorIO):
query: InterleavedContent,
params: Optional[Dict[str, Any]] = None,
) -> QueryChunksResponse:
logcat.debug("core", f"VectorIORouter.query_chunks: {vector_db_id}")
return await self.routing_table.get_provider_impl(vector_db_id).query_chunks(vector_db_id, query, params)
@ -111,12 +121,15 @@ class InferenceRouter(Inference):
self,
routing_table: RoutingTable,
) -> None:
logcat.debug("core", "Initializing InferenceRouter")
self.routing_table = routing_table
async def initialize(self) -> None:
logcat.debug("core", "InferenceRouter.initialize")
pass
async def shutdown(self) -> None:
logcat.debug("core", "InferenceRouter.shutdown")
pass
async def register_model(
@ -127,6 +140,10 @@ class InferenceRouter(Inference):
metadata: Optional[Dict[str, Any]] = None,
model_type: Optional[ModelType] = None,
) -> None:
logcat.debug(
"core",
f"InferenceRouter.register_model: {model_id=} {provider_model_id=} {provider_id=} {metadata=} {model_type=}",
)
await self.routing_table.register_model(model_id, provider_model_id, provider_id, metadata, model_type)
async def chat_completion(
@ -142,6 +159,10 @@ class InferenceRouter(Inference):
logprobs: Optional[LogProbConfig] = None,
tool_config: Optional[ToolConfig] = None,
) -> AsyncGenerator:
logcat.debug(
"core",
f"InferenceRouter.chat_completion: {model_id=}, {stream=}, {messages=}, {tools=}, {tool_config=}, {response_format=}",
)
model = await self.routing_table.get_model(model_id)
if model is None:
raise ValueError(f"Model '{model_id}' not found")
@ -203,6 +224,10 @@ class InferenceRouter(Inference):
stream: Optional[bool] = False,
logprobs: Optional[LogProbConfig] = None,
) -> AsyncGenerator:
logcat.debug(
"core",
f"InferenceRouter.completion: {model_id=}, {stream=}, {content=}, {sampling_params=}, {response_format=}",
)
model = await self.routing_table.get_model(model_id)
if model is None:
raise ValueError(f"Model '{model_id}' not found")
@ -230,6 +255,7 @@ class InferenceRouter(Inference):
output_dimension: Optional[int] = None,
task_type: Optional[EmbeddingTaskType] = None,
) -> EmbeddingsResponse:
logcat.debug("core", f"InferenceRouter.embeddings: {model_id}")
model = await self.routing_table.get_model(model_id)
if model is None:
raise ValueError(f"Model '{model_id}' not found")
@ -249,12 +275,15 @@ class SafetyRouter(Safety):
self,
routing_table: RoutingTable,
) -> None:
logcat.debug("core", "Initializing SafetyRouter")
self.routing_table = routing_table
async def initialize(self) -> None:
logcat.debug("core", "SafetyRouter.initialize")
pass
async def shutdown(self) -> None:
logcat.debug("core", "SafetyRouter.shutdown")
pass
async def register_shield(
@ -264,6 +293,7 @@ class SafetyRouter(Safety):
provider_id: Optional[str] = None,
params: Optional[Dict[str, Any]] = None,
) -> Shield:
logcat.debug("core", f"SafetyRouter.register_shield: {shield_id}")
return await self.routing_table.register_shield(shield_id, provider_shield_id, provider_id, params)
async def run_shield(
@ -272,6 +302,7 @@ class SafetyRouter(Safety):
messages: List[Message],
params: Dict[str, Any] = None,
) -> RunShieldResponse:
logcat.debug("core", f"SafetyRouter.run_shield: {shield_id}")
return await self.routing_table.get_provider_impl(shield_id).run_shield(
shield_id=shield_id,
messages=messages,
@ -284,12 +315,15 @@ class DatasetIORouter(DatasetIO):
self,
routing_table: RoutingTable,
) -> None:
logcat.debug("core", "Initializing DatasetIORouter")
self.routing_table = routing_table
async def initialize(self) -> None:
logcat.debug("core", "DatasetIORouter.initialize")
pass
async def shutdown(self) -> None:
logcat.debug("core", "DatasetIORouter.shutdown")
pass
async def get_rows_paginated(
@ -299,6 +333,7 @@ class DatasetIORouter(DatasetIO):
page_token: Optional[str] = None,
filter_condition: Optional[str] = None,
) -> PaginatedRowsResult:
logcat.debug("core", f"DatasetIORouter.get_rows_paginated: {dataset_id}, rows_in_page={rows_in_page}")
return await self.routing_table.get_provider_impl(dataset_id).get_rows_paginated(
dataset_id=dataset_id,
rows_in_page=rows_in_page,
@ -307,6 +342,7 @@ class DatasetIORouter(DatasetIO):
)
async def append_rows(self, dataset_id: str, rows: List[Dict[str, Any]]) -> None:
logcat.debug("core", f"DatasetIORouter.append_rows: {dataset_id}, {len(rows)} rows")
return await self.routing_table.get_provider_impl(dataset_id).append_rows(
dataset_id=dataset_id,
rows=rows,
@ -318,12 +354,15 @@ class ScoringRouter(Scoring):
self,
routing_table: RoutingTable,
) -> None:
logcat.debug("core", "Initializing ScoringRouter")
self.routing_table = routing_table
async def initialize(self) -> None:
logcat.debug("core", "ScoringRouter.initialize")
pass
async def shutdown(self) -> None:
logcat.debug("core", "ScoringRouter.shutdown")
pass
async def score_batch(
@ -332,6 +371,7 @@ class ScoringRouter(Scoring):
scoring_functions: Dict[str, Optional[ScoringFnParams]] = None,
save_results_dataset: bool = False,
) -> ScoreBatchResponse:
logcat.debug("core", f"ScoringRouter.score_batch: {dataset_id}")
res = {}
for fn_identifier in scoring_functions.keys():
score_response = await self.routing_table.get_provider_impl(fn_identifier).score_batch(
@ -352,6 +392,7 @@ class ScoringRouter(Scoring):
input_rows: List[Dict[str, Any]],
scoring_functions: Dict[str, Optional[ScoringFnParams]] = None,
) -> ScoreResponse:
logcat.debug("core", f"ScoringRouter.score: {len(input_rows)} rows, {len(scoring_functions)} functions")
res = {}
# look up and map each scoring function to its provider impl
for fn_identifier in scoring_functions.keys():
@ -369,12 +410,15 @@ class EvalRouter(Eval):
self,
routing_table: RoutingTable,
) -> None:
logcat.debug("core", "Initializing EvalRouter")
self.routing_table = routing_table
async def initialize(self) -> None:
logcat.debug("core", "EvalRouter.initialize")
pass
async def shutdown(self) -> None:
logcat.debug("core", "EvalRouter.shutdown")
pass
async def run_eval(
@ -382,6 +426,7 @@ class EvalRouter(Eval):
benchmark_id: str,
task_config: BenchmarkConfig,
) -> Job:
logcat.debug("core", f"EvalRouter.run_eval: {benchmark_id}")
return await self.routing_table.get_provider_impl(benchmark_id).run_eval(
benchmark_id=benchmark_id,
task_config=task_config,
@ -394,6 +439,7 @@ class EvalRouter(Eval):
scoring_functions: List[str],
task_config: BenchmarkConfig,
) -> EvaluateResponse:
logcat.debug("core", f"EvalRouter.evaluate_rows: {benchmark_id}, {len(input_rows)} rows")
return await self.routing_table.get_provider_impl(benchmark_id).evaluate_rows(
benchmark_id=benchmark_id,
input_rows=input_rows,
@ -406,6 +452,7 @@ class EvalRouter(Eval):
benchmark_id: str,
job_id: str,
) -> Optional[JobStatus]:
logcat.debug("core", f"EvalRouter.job_status: {benchmark_id}, {job_id}")
return await self.routing_table.get_provider_impl(benchmark_id).job_status(benchmark_id, job_id)
async def job_cancel(
@ -413,6 +460,7 @@ class EvalRouter(Eval):
benchmark_id: str,
job_id: str,
) -> None:
logcat.debug("core", f"EvalRouter.job_cancel: {benchmark_id}, {job_id}")
await self.routing_table.get_provider_impl(benchmark_id).job_cancel(
benchmark_id,
job_id,
@ -423,6 +471,7 @@ class EvalRouter(Eval):
benchmark_id: str,
job_id: str,
) -> EvaluateResponse:
logcat.debug("core", f"EvalRouter.job_result: {benchmark_id}, {job_id}")
return await self.routing_table.get_provider_impl(benchmark_id).job_result(
benchmark_id,
job_id,
@ -435,6 +484,7 @@ class ToolRuntimeRouter(ToolRuntime):
self,
routing_table: RoutingTable,
) -> None:
logcat.debug("core", "Initializing ToolRuntimeRouter.RagToolImpl")
self.routing_table = routing_table
async def query(
@ -443,6 +493,7 @@ class ToolRuntimeRouter(ToolRuntime):
vector_db_ids: List[str],
query_config: Optional[RAGQueryConfig] = None,
) -> RAGQueryResult:
logcat.debug("core", f"ToolRuntimeRouter.RagToolImpl.query: {vector_db_ids}")
return await self.routing_table.get_provider_impl("knowledge_search").query(
content, vector_db_ids, query_config
)
@ -453,6 +504,10 @@ class ToolRuntimeRouter(ToolRuntime):
vector_db_id: str,
chunk_size_in_tokens: int = 512,
) -> None:
logcat.debug(
"core",
f"ToolRuntimeRouter.RagToolImpl.insert: {vector_db_id}, {len(documents)} documents, chunk_size={chunk_size_in_tokens}",
)
return await self.routing_table.get_provider_impl("insert_into_memory").insert(
documents, vector_db_id, chunk_size_in_tokens
)
@ -461,6 +516,7 @@ class ToolRuntimeRouter(ToolRuntime):
self,
routing_table: RoutingTable,
) -> None:
logcat.debug("core", "Initializing ToolRuntimeRouter")
self.routing_table = routing_table
# HACK ALERT this should be in sync with "get_all_api_endpoints()"
@ -469,12 +525,15 @@ class ToolRuntimeRouter(ToolRuntime):
setattr(self, f"rag_tool.{method}", getattr(self.rag_tool, method))
async def initialize(self) -> None:
logcat.debug("core", "ToolRuntimeRouter.initialize")
pass
async def shutdown(self) -> None:
logcat.debug("core", "ToolRuntimeRouter.shutdown")
pass
async def invoke_tool(self, tool_name: str, kwargs: Dict[str, Any]) -> Any:
logcat.debug("core", f"ToolRuntimeRouter.invoke_tool: {tool_name}")
return await self.routing_table.get_provider_impl(tool_name).invoke_tool(
tool_name=tool_name,
kwargs=kwargs,
@ -483,4 +542,5 @@ class ToolRuntimeRouter(ToolRuntime):
async def list_runtime_tools(
self, tool_group_id: Optional[str] = None, mcp_endpoint: Optional[URL] = None
) -> List[ToolDef]:
logcat.debug("core", f"ToolRuntimeRouter.list_runtime_tools: {tool_group_id}")
return await self.routing_table.get_provider_impl(tool_group_id).list_tools(tool_group_id, mcp_endpoint)