youngkingdom/ragflow
1
0
Fork 0
mirror of https://github.com/infiniflow/ragflow.git synced 2026-05-28 03:33:05 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix: Langfuse chat observation (#15026)

Browse Source 
### What problem does this PR solve?

Closes #15025

Langfuse-enabled `dialog_service.async_chat()` regressed to
`langfuse_tracer.start_generation(...)` after the earlier Langfuse v4
migration. Langfuse v4 uses `start_observation(as_type="generation")`,
so the remaining `start_generation` call can fail when chat tracing is
enabled.

This restores the migrated `start_observation(as_type="generation")`
call for chat observations while preserving the existing trace context,
model, input payload, and update/end flow. It also adds a regression
test with a fake Langfuse v4-style client that exposes
`start_observation()` but not `start_generation()`.

### Type of change

- [x] Bug Fix (non-breaking change which fixes an issue)

### Tests

- `.venv/bin/pytest
test/unit_test/api/db/services/test_dialog_service_final_answer.py -q`
- `.venv/bin/ruff check api/db/services/dialog_service.py
test/unit_test/api/db/services/test_dialog_service_final_answer.py`
This commit is contained in:
bitloi
2026-05-20 04:01:19 -03:00
committed by GitHub
parent 1ed8a118cf
commit 6499bce2a6
2 changed files with 180 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
api/db/services/dialog_service.py
View File

@ -572,6 +572,7 @@ async def async_chat(dialog, messages, stream=True, **kwargs):
check_llm_ts = timer()
langfuse_tracer = None
langfuse_generation = None
trace_context = {}
langfuse_keys = TenantLangfuseService.filter_by_tenant(tenant_id=dialog.tenant_id)
if langfuse_keys:
@ -783,7 +784,7 @@ async def async_chat(dialog, messages, stream=True, **kwargs):
gen_conf["max_tokens"] = min(gen_conf["max_tokens"], max_tokens - used_token_count)
async def decorate_answer(answer):
nonlocal embd_mdl, prompt_config, knowledges, kwargs, kbinfos, prompt, retrieval_ts, questions, langfuse_tracer
nonlocal embd_mdl, prompt_config, knowledges, kwargs, kbinfos, prompt, retrieval_ts, questions, langfuse_generation
refs = []
ans = answer.split("</think>")
@ -855,8 +856,8 @@ async def async_chat(dialog, messages, stream=True, **kwargs):
f" - Token speed: {int(tk_num / (generate_result_time_cost / 1000.0))}/s"
)
# Add a condition check to call the end method only if langfuse_tracer exists
if langfuse_tracer and "langfuse_generation" in locals():
# Add a condition check to call the end method only if langfuse_generation exists
if langfuse_generation is not None:
langfuse_output = "\n" + re.sub(r"^.*?(### Query:.*)", r"\1", prompt, flags=re.DOTALL)
langfuse_output = {"time_elapsed:": re.sub(r"\n", " \n", langfuse_output), "created_at": time.time()}
langfuse_generation.update(
@ -872,9 +873,18 @@ async def async_chat(dialog, messages, stream=True, **kwargs):
return {"answer": think + answer, "reference": refs, "prompt": re.sub(r"\n", " \n", prompt), "created_at": time.time()}
if langfuse_tracer:
langfuse_generation = langfuse_tracer.start_generation(
trace_context=trace_context, name="chat", model=llm_model_config["llm_name"], input={"prompt": prompt, "prompt4citation": prompt4citation, "messages": msg}
)
try:
langfuse_generation = langfuse_tracer.start_observation(
as_type="generation",
trace_context=trace_context,
name="chat",
model=llm_model_config["llm_name"],
input={"prompt": prompt, "prompt4citation": prompt4citation, "messages": msg},
)
except Exception as e: # noqa: BLE001 - tracing must not break chat flow
logger.warning("Langfuse start_observation failed; continuing without tracing: %s", e)
langfuse_tracer = None
langfuse_generation = None
if stream:
if llm_type == "chat":