adiuvAI/api
1
0
Fork 0
Code Issues 1 Pull Requests Actions Packages Projects Releases Wiki Activity

Add Langfuse observability: traces, prompt management, prompt-to-generation linking

Browse Source 
- New app/core/langfuse_client.py: lazy singleton client, get_prompt_or_fallback()
  helper (returns raw template + prompt obj for linking), extract_usage() for token
  counts. No-ops when LANGFUSE_* env vars are not set.
- deep_agent.py: home-agent and floating-agent runs wrapped in spans; each ainvoke
  wrapped in a generation with model/input/output/usage; prompts fetched from
  Langfuse (adiuva-home-agent, adiuva-floating-agent, adiuva-floating-classifier)
  with hardcoded fallback.
- agent_runner.py: step1-classifier and step2-processor LLM calls traced; batch
  agent _run_agent_with_tools spans + generations; cloud-processor included.
  Prompts: adiuva-step1-classifier, adiuva-step2-processor, adiuva-cloud-processor.
- agent_setup.py: journey-setup span + generation per ainvoke; prompt_obj stored
  on JourneySession and reused across turns. Prompt: journey_system.
- settings.py: LANGFUSE_SECRET_KEY, LANGFUSE_PUBLIC_KEY, LANGFUSE_HOST added.
- .env.example: Langfuse section with EU/US/self-hosted host comments.
- requirements.txt: langfuse>=2.0.0.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This commit is contained in:
Roberto Musso
2026-04-07 00:19:20 +02:00
parent 552b8eb305
commit 1ce1d492b0
7 changed files with 455 additions and 78 deletions
Download Patch File Download Diff File Expand all files Collapse all files

121
app/api/routes/agent_setup.py
View File

@@ -31,6 +31,8 @@ from typing import Any
from langchain_core.messages import AIMessage, HumanMessage, SystemMessage, ToolMessage
from app.agents.filesystem_agent import FILESYSTEM_TOOLS
from app.config.settings import settings
from app.core.langfuse_client import extract_usage, get_langfuse, get_prompt_or_fallback
from app.core.llm import get_llm
logger = logging.getLogger(__name__)
@@ -62,6 +64,7 @@ class JourneySession:
data_types: list[str]
history: list[dict[str, Any]] = field(default_factory=list)
system_prompt: str = ""
langfuse_prompt: Any = None
created_at: float = field(default_factory=time.monotonic)
def is_expired(self) -> bool:
@@ -146,20 +149,25 @@ def _build_system_prompt(
directory: str,
data_types: list[str],
existing_template: str | None = None,
) -> str:
) -> tuple[str, Any]:
"""Return ``(compiled_system_prompt, langfuse_prompt_obj_or_None)``."""
existing_section = (
f"\nThe user already has the following prompt_template — refine it based on their answers:\n"
f"---\n{existing_template}\n---\n"
if existing_template
else ""
)
return _SYSTEM_PROMPT_TEMPLATE.format(
template, prompt_obj = get_prompt_or_fallback(
"journey_system", _SYSTEM_PROMPT_TEMPLATE
)
compiled = template.format(
directory=directory,
data_types=", ".join(data_types),
template_start=_TEMPLATE_START,
template_end=_TEMPLATE_END,
existing_section=existing_section,
)
return compiled, prompt_obj
# ── Template extraction ───────────────────────────────────────────────────
@@ -199,12 +207,17 @@ async def _call_llm_with_tools(
system_prompt: str,
history: list[dict[str, Any]],
tools: list[Any],
*,
user_id: str = "",
session_id: str = "",
langfuse_prompt: Any = None,
) -> str:
"""Build LangChain messages from history and invoke the LLM with tools.
Handles tool-calling loops: if the LLM calls tools, execute them and
continue until a final text response is produced.
"""
lf = get_langfuse()
messages: list[Any] = [SystemMessage(content=system_prompt)]
for turn in history:
if turn["role"] == "user":
@@ -216,38 +229,76 @@ async def _call_llm_with_tools(
llm_with_tools = llm.bind_tools(tools)
tool_map = {tool_def.name: tool_def for tool_def in tools}
for _ in range(_MAX_TOOL_STEPS):
response: AIMessage = await llm_with_tools.ainvoke(messages)
messages.append(response)
_span_ctx = (
lf.start_as_current_observation(
as_type="span",
name="journey-setup",
user_id=user_id or None,
session_id=session_id or None,
input=history[-1]["content"] if history else "",
)
if lf else None
)
_span = _span_ctx.__enter__() if _span_ctx else None
if not response.tool_calls:
return _as_text(response.content)
for call in response.tool_calls:
call_name = str(call.get("name", ""))
call_args = call.get("args", {})
logger.info(
"agent_setup: journey tool_call name=%s args=%s",
call_name,
json.dumps(call_args, ensure_ascii=True)[:500],
try:
for _ in range(_MAX_TOOL_STEPS):
_gen_ctx = (
lf.start_as_current_observation(
as_type="generation",
name="journey-setup-llm",
model=settings.LLM_MODEL,
prompt=langfuse_prompt,
input=messages,
)
if lf else None
)
_gen = _gen_ctx.__enter__() if _gen_ctx else None
response: AIMessage = await llm_with_tools.ainvoke(messages)
if _gen_ctx:
_gen.update(output=_as_text(response.content), usage=extract_usage(response))
_gen_ctx.__exit__(None, None, None)
tool_fn = tool_map.get(call_name)
if tool_fn is None:
tool_output = f"Unknown tool: {call_name}"
else:
tool_output = await tool_fn.ainvoke(call_args)
messages.append(response)
logger.info(
"agent_setup: journey tool_result name=%s output=%s",
call_name,
str(tool_output)[:800],
)
messages.append(ToolMessage(content=str(tool_output), tool_call_id=call["id"]))
if not response.tool_calls:
if _span:
_span.update(output=_as_text(response.content))
return _as_text(response.content)
# Fallback: exceeded max steps.
final = await llm.ainvoke(messages)
return _as_text(final.content)
for call in response.tool_calls:
call_name = str(call.get("name", ""))
call_args = call.get("args", {})
logger.info(
"agent_setup: journey tool_call name=%s args=%s",
call_name,
json.dumps(call_args, ensure_ascii=True)[:500],
)
tool_fn = tool_map.get(call_name)
if tool_fn is None:
tool_output = f"Unknown tool: {call_name}"
else:
tool_output = await tool_fn.ainvoke(call_args)
logger.info(
"agent_setup: journey tool_result name=%s output=%s",
call_name,
str(tool_output)[:800],
)
messages.append(ToolMessage(content=str(tool_output), tool_call_id=call["id"]))
# Fallback: exceeded max steps.
final = await llm.ainvoke(messages)
final_text = _as_text(final.content)
if _span:
_span.update(output=final_text)
return final_text
finally:
if _span_ctx:
_span_ctx.__exit__(None, None, None)
if lf:
lf.flush()
# ── Journey handlers (called from device_ws.py) ──────────────────────────
@@ -270,7 +321,7 @@ async def handle_journey_start(
# Use the session_id provided by the FE so the reply matches the
# listener key; fall back to a generated one if absent.
session_id = frame.get("session_id") or str(uuid.uuid4())
system_prompt = _build_system_prompt(directory, data_types, existing_template)
system_prompt, langfuse_prompt = _build_system_prompt(directory, data_types, existing_template)
session = JourneySession(
session_id=session_id,
@@ -279,6 +330,7 @@ async def handle_journey_start(
directory=directory,
data_types=data_types,
system_prompt=system_prompt,
langfuse_prompt=langfuse_prompt,
)
# The LLM will explore the directory using FILESYSTEM_TOOLS via the
@@ -292,6 +344,9 @@ async def handle_journey_start(
system_prompt=system_prompt,
history=seed_history,
tools=list(FILESYSTEM_TOOLS),
user_id=user_id,
session_id=session_id,
langfuse_prompt=langfuse_prompt,
)
session.history.extend(seed_history)
@@ -356,6 +411,9 @@ async def handle_journey_message(
system_prompt=session.system_prompt,
history=session.history,
tools=list(FILESYSTEM_TOOLS),
user_id=session.user_id,
session_id=session_id,
langfuse_prompt=session.langfuse_prompt,
)
session.history.append({"role": "assistant", "content": ai_reply})
@@ -379,6 +437,9 @@ async def handle_journey_message(
system_prompt=session.system_prompt,
history=session.history,
tools=list(FILESYSTEM_TOOLS),
user_id=session.user_id,
session_id=session_id,
langfuse_prompt=session.langfuse_prompt,
)
session.history.append({"role": "assistant", "content": nudge_reply})