Adds a Neo4j service to docker-compose so `docker compose up -d` works
on a clean checkout, and unhardcodes Graphiti's LLM/embedder so the
documented default provider (Qwen via Dashscope) actually works.
- docker-compose: neo4j:5-community service with cypher-shell
healthcheck, named volumes, and `depends_on: service_healthy` on the
app container; in-Docker NEO4J_URI override leaves the host-mode
default untouched.
- Config: new GRAPHITI_LLM_PROVIDER (openai|gemini, default openai) plus
optional EMBEDDING_API_KEY / EMBEDDING_BASE_URL that fall back to the
chat LLM credentials.
- graphiti_adapter: provider switch inside the singleton factory with
lazy per-provider imports; Gemini path is preserved exactly. The
no-op `_GeminiReranker` becomes a provider-agnostic
`_PassthroughReranker`, still injected explicitly so Graphiti does
not fall back to its OpenAI-only default reranker.
- Drop the ignored `reranker=` kwarg from `_GraphNamespace.search` and
the misleading callers in `zep_tools.py` and
`oasis_profile_generator.py`.
- Refresh `.env.example` to mirror the README env section.
Spec, requirements, and design under
`.kiro/specs/graphiti-neo4j-finalize/`.
Closes#1
Background threads (graph building, simulation prep, report generation,
profile generation) now inherit the requesting user's locale preference.
Previously these fell back to 'zh' because Flask request context was
unavailable in spawned threads.
Ensure poster_type stays PascalCase English and stance stays English enum
values regardless of language setting. Only natural language fields follow
the user's language preference.
The language instruction was causing LLM to change entity/relation naming
conventions. Now explicitly enforce PascalCase/UPPER_SNAKE_CASE for technical
identifiers while only applying language preference to description fields.
- Changed error messages to reflect the new configuration requirement for Neo4j.
- Ensured consistent handling of missing credentials across multiple functions.
Replaces the paid, rate-limited Zep Cloud service with Graphiti (graphiti-core
0.11.6) backed by a local Neo4j instance — free, unlimited, and self-hosted.
Key changes:
- Add GraphitiAdapter: drop-in Zep-compatible wrapper around Graphiti with a
persistent event-loop thread to avoid asyncio/Neo4j driver conflicts
- Switch LLM client to native GeminiClient + GeminiEmbedder (text-embedding-004
fails on Gemini compat endpoint; use google-genai SDK directly)
- Add _GeminiReranker passthrough replacing OpenAIRerankerClient (which
hardcodes gpt-4.1-nano and uses logprobs unsupported by Gemini)
- Fix Cypher queries: use s.uuid/t.uuid for edge source/target instead of
r.source_node_uuid (null property in Graphiti's schema)
- Add ontology-based entity type classifier (_classify_entity_type) so nodes
get colored by type in the D3 graph visualization instead of all being Entity
- Apply classifier in ZepEntityReader so filter_defined_entities finds entities
(previously 0 personas loaded because all labels were ['Entity'])
- Add startup recovery: auto-mark graph_building projects as graph_completed
on backend restart if Neo4j already has their data
- Add resume capability to graph build: skip already-processed episodes after
a restart (断点续传)
- Add non-blocking graph data cache with background refresh in graph.py
- Add EMBEDDING_MODEL config (default: gemini-embedding-001 for Gemini users)
- Add CLAUDE.md with project architecture and dev commands
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Implemented `_get_report_id_for_simulation` to find the most recent report ID associated with a simulation ID by scanning the reports directory.
- Updated `get_simulation_history` to include the retrieved report ID in the response, enhancing the simulation data returned to the client.
- Updated simulation history retrieval to read project details directly from the Simulation file.
- Improved simulation configuration handling by reading simulation requirements from JSON.
- Added project file listing to the simulation history, displaying up to three associated files.
- Refined card layout in HistoryDatabase.vue to accommodate new file display features and improved responsiveness.
- Decreased the maximum tool calls per section from 8 to 5.
- Reduced the maximum iterations in the ReACT loop from 8 to 5, streamlining the report generation process.
- Reduced maximum tool calls per chat from 5 to 2 for improved efficiency.
- Simplified system prompt to focus on concise responses and report content.
- Implemented report content retrieval with length limitation to prevent context overflow.
- Adjusted tool call execution to limit to one call per iteration, enhancing clarity in responses.
- Updated user message prompts to encourage concise answers based on retrieved data.
- Increased the maximum tool calls per section from 4 to 8, enhancing the agent's capabilities.
- Raised the maximum reflection rounds from 2 to 3 to allow for deeper analysis.
- Adjusted the maximum tool calls per chat from 3 to 5 for improved interaction.
- Expanded the maximum agents for interviews from 5 to 20, facilitating more comprehensive data gathering.
- Increased the maximum iterations for ReACT loops from 5 to 8 and from 3 to 5 in different contexts, optimizing the report generation process.
- Updated the `to_text` method in the `PanoramaResult` class to provide complete outputs for current facts, historical facts, and involved entities, improving data visibility.
- Modified the `to_text` method in the `AgentInterview` class to display the full agent bio without truncation.
- Adjusted the `ZepToolsService` class to retrieve all related entity details and facts without limiting the output, ensuring comprehensive data representation.
- Renamed log_section_complete to log_section_content to better reflect its purpose, and added is_subsection parameter for improved logging of subsection content.
- Introduced log_section_full_complete method to log the completion of entire sections, including all subsections, enhancing tracking of report generation status.
- Adjusted maximum tool call limits for sections and chats to optimize performance during report generation.
- Updated system prompts and user prompts in the ReportAgent class to clarify the report's focus on future predictions rather than current analysis.
- Enhanced the Step3Simulation and Step4Report components for improved user experience, including UI updates and better handling of report generation states.
- Updated the AgentInterview class to display the full agent bio, truncating only if it exceeds 1000 characters for better readability.
- Enhanced the Step4Report component to include structured display for tool results, allowing users to toggle between raw and structured views for various tools, improving user experience and clarity.
- Introduced new components for parsing and displaying results from different tools, including InsightForge, PanoramaSearch, InterviewAgents, and QuickSearch, providing a comprehensive view of the data.
- Introduced a unique report ID generation mechanism to enhance tracking and management of reports.
- Implemented detailed logging for the report generation process, including agent actions, planning stages, and tool calls, improving traceability and debugging.
- Added new API endpoints for retrieving agent and console logs, allowing users to access detailed execution logs and console outputs during report generation.
- Enhanced the frontend GraphPanel component with a notification for users when simulations finish, improving user experience and feedback.
- Introduced strict formatting guidelines for chapter content, prohibiting the use of Markdown headers and emphasizing the use of bold text for section titles.
- Implemented a new method to save chapters along with their subsections into a single file, streamlining the report structure.
- Added content cleaning functionality to remove duplicate titles and ensure proper formatting before saving.
- Enhanced the report assembly process to include post-processing for title management and improved readability.
Dominik Seemann
stg
BaiFu
ghostubborn
Abhishek Yadav