Add FalkorDB Knowledge Graph with GraphRAG for LEGI, JADE, and BOFIP

[Copilot]

Adds FalkorDB as a graph database to build a Knowledge Graph over processed legal data (LEGI, JADE, BOFIP), with GraphRAG SDK integration for entity-aware Q&A.

Design decision: Option B — parallel population during processing

Graph nodes and relationships are written to FalkorDB in the same pass as PostgreSQL inserts, while the raw XML is being parsed. This avoids a second ETL pass and preserves structural metadata (e.g. LIENS cross-references) only available at parse time. All graph ops are wrapped in try/except — FalkorDB unavailability never interrupts the main pipeline.

Infrastructure

• docker-compose.yml: Add falkordb service (falkordb/falkordb:v4.4.4) with healthcheck, named volume, port 6379, and Airflow depends_on.
• .env.example / config/config.py / config/__init__.py: Add FALKORDB_HOST, FALKORDB_PORT, FALKORDB_USERNAME, FALKORDB_PASSWORD, FALKORDB_GRAPH_NAME.
• pyproject.toml: Add falkordb==1.6.0 and graphrag-sdk==0.8.2.

Graph ontology (database/graph_manage.py)

Node	Source
LegalText	LEGI articles
JudicialDecision	JADE decisions
TaxGuidance	BOFIP guidance
LegalCode	Code civil, Code pénal, …
Ministry	Issuing ministries
Jurisdiction	Administrative courts
Chunk	Text chunks (GraphRAG)

Relationships: BELONGS_TO_CODE, ISSUED_BY, REFERENCES (from LIENS metadata), DECIDED_BY, PART_OF.

Key functions

• init_graph_schema() — creates FalkorDB indexes
• upsert_{legi,jade,bofip}_{node,chunk}() — parallel-population upserts
• populate_graph_from_postgres() — streaming back-fill from existing PostgreSQL data (server-side cursors)
• build_graphrag_knowledge_graph() — instantiates a graphrag_sdk.KnowledgeGraph with the full ontology for entity extraction / Q&A

Processing pipeline (files_processing.py)

Graph upserts added after each PostgreSQL insert for LEGI and JADE handlers. Added a new BOFIPTEXT*.xml handler (matching the existing DILA XML pattern) and a "bofip" type branch in process_data() to handle the BOFIP tgz archives.

# After PostgreSQL insert in LEGI handler
upsert_legi_node(doc_id=cid, nature=nature, category=category,
                 ministry=ministry, links=links, ...)
for chunk_id, *_ in data_to_insert:
    upsert_legi_chunk(chunk_id=chunk_id, doc_id=cid)

BOFIP PostgreSQL table

Added BOFIP table schema to database_manage.py (mirrors the LEGI/JADE pattern with chunk_id, doc_id, nature, category, title, date, text, chunk_text, vector embedding column).

Airflow DAG (airflow_config/dags/knowledge_graph.py)

New KNOWLEDGE_GRAPH DAG with two tasks: init_graph_schema → populate_graph_from_postgres, for bootstrapping or back-filling existing deployments.

Original prompt

---

This section details on the original issue you should resolve

<issue_title>Falkdor DB setup with GraphRAG and Knowledge graph</issue_title>
<issue_description>This is a critical setup so be accurate and analyze well before.
We are going to add FalkorDB to the docker compose first.

Then based on a deep understanding of the processing of JADE, BOFIP and LEGIT, we will add nodes and relationships to a FalkorDB graph.

We then aim to build an Knowledge graph based on the data we processed. Here you need a judment call: Either we leverage data already processed from the database , OR, we populate the Graph when we are processing the raw data, in parallel to populating the postgresDB.

Once you have the onlogy and graph, you will design a Knowledge graph. I suggest this SDK, but you are free
https://github.com/FalkorDB/GraphRAG-SDK/

Take all your time for the design.</issue_description>

Comments on the Issue (you are @copilot in this section)

• Fixes Falkdor DB setup with GraphRAG and Knowledge graph #21

---

📍 Connect Copilot coding agent with Jira, Azure Boards or Linear to delegate work to Copilot in one click without leaving your project management tool.

[yawo]

Close this

[Copilot]

Close this

I don't have the ability to close pull requests directly. You can close it from the GitHub UI by clicking "Close pull request" at the bottom of the PR page.