Agent refactor#343
Open
renan-souza wants to merge 43 commits into
Open
Conversation
…s to new paths
- Deleted flowcept_agent.py, flowcept_ctx_manager.py, entire tools/ directory,
and old prompts (general_prompts.py, in_memory_query_prompts.py,
workflow_query_prompts.py) — all replaced by the new mcp_server/context_manager/
mcp_tools/prompts structure
- Moved ROUTING_PROMPT and SMALL_TALK_PROMPT from deleted general_prompts.py into
base_prompts.py; updated session_tools.py import accordingly
- Updated all six stale test imports across agent_tests.py and
test_webservice_integration.py to point at the new module paths
(mcp_tools/, data_query_tools/, context_manager, mcp_prompts)
Wire the PROV-AGENT model (arXiv:2508.02866) throughout the Flowcept agent stack so every LLM call, MCP tool execution, and LangGraph chat turn is captured as a typed provenance task. - Replace bare "llm_task" / "agent_task" strings with PROV_AGENT.AI_MODEL_INVOCATION / PROV_AGENT.AGENT_TOOL enum values in FlowceptLLM, agent_flowcept_task, and context_manager.py - Add @agent_flowcept_task(subtype=PROV_AGENT.AGENT_TOOL) to all MCP tool wrappers (db_query_mcp_tools, in_memory_task_query_mcp_tools, report_tools, session_tools) - Wrap each run_chat / LangGraph execution in Flowcept(workflow_name="langgraph_chat", start_persistence=False, save_workflow=True) so every chat turn has its own workflow_id; call_model uses FlowceptLLM, call_tools uses FlowceptTask — both inherit Flowcept.current_workflow_id automatically - Fix FlowceptLLM._format_messages to handle BaseMessage objects (needed for LangGraph state) - Add buffer-None guard to agent_flowcept_task so the decorator skips gracefully when MQ is not started - Expand PROV_AGENT enum docstrings with W3C PROV mapping and paper citation; improve task_object.py subtype docstring with PROV-AGENT examples - Update docs/schemas.rst (paper citation, vocabulary table, wasInformedBy explanation) and agents/README.md (new PROV-AGENT Instrumentation section) - 11 new TestProvAgentInstrumentation structural tests; all pass
Move MCP context handling under the agent MCP package, retire stale prompt-handler paths, and clarify agent/webservice separation docs. Replace fake webservice API tests with real integration assertions, and update dashboard routes to use the real dashboard DAO interface.
- Flowcept.services_alive() now returns ServicesAliveResult(dict) — a dict
subclass with __bool__, so all existing callers are unaffected while
the health endpoint can expose per-service status
- DBAPI.db_liveness_tests() added to test Mongo and LMDB independently
(fixes the bug where both-enabled would only test whichever DAO won
get_instance()); flowcept_controller calls DBAPI — no DAO imports leak
out of db_api.py (SoC fix)
- LLM liveness folded into services_alive() when AGENT_CHAT_ENABLED and
api_key is non-placeholder — settings-driven, no new method
- /api/v1/health/ready returns {"status", "services": {…}} with per-service
breakdown; HTTP 503 when any service is unavailable
- T18: session-scoped gridsearch_run_data fixture (conftest.py), cosine-
similarity scorer (tests/test_utils/test_llm_utils.py), 21-case
chat_query_tests.yaml with deterministic expected values, and two new
tests (test_chat_endpoint_real_llm_db_queries,
test_chat_endpoint_real_llm_df_queries); 18/20 webservice tests pass,
2 blocked on orchestrator recursion limit, still investigating.
Changes: - Missing settings now use built-in DEFAULT_SETTINGS, not sample_settings.yaml, so offline buffer dumping is enabled by default. - Flowcept.generate_report() with no input now reads the configured default buffer file. - examples/start_here.py no longer requires flowcept --init-settings. - rich is now a base dependency so print_markdown=True works after pip install flowcept. - run-tests.yml now tests examples/start_here.py after deleting settings, with no init step.
…ocalhost, 127.0.0.1, ::1}:
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.