SUSTAINABILITY INTELLIGENCE · BRSR · HYBRID RAG

Evidence before generation.

POSTGRESQL · PGVECTOR · HYBRID SEARCH

Ask real ESG disclosures. Trace every answer to the page.

The system embeds your query with Supabase gte-small, fuses semantic vector similarity with PostgreSQL full-text search, and sends only the retrieved evidence to the generation layer.

HOW TO USE THIS TOOL

From question to evidence in three steps.

Use ESG RAG Intelligence to interrogate the indexed BRSR framework, BRSR Core guidance and corporate ESG disclosures. The tool is designed for evidence-led research, not as a replacement for the underlying reports.

01

Ask a specific question

Ask about an ESG topic, regulatory requirement, company disclosure or comparison. For example: “Compare Infosys and Reliance on climate-related performance.”

02

Refine the research

Use the source and ESG-theme filters when you want to narrow the search. For comparisons, name both companies and ask specifically about reporting year, methodology, scope or boundary differences.

03

Verify the evidence

Material claims in the answer cite evidence IDs such as [E1] and [E2]. Review the Retrieval Trace, check the PDF page and use “source ↗” to inspect the original disclosure.

Research tip. If the retrieved evidence is insufficient, broaden the filters or reformulate the question. Absence from the retrieved evidence should not automatically be interpreted as absence from the full source document.
QUERY

ESG research question

HYBRID RRF
GROUNDED RESPONSE

Executive analysis

Ready
The generated response will appear here. Claims must cite retrieved evidence IDs such as [E1] and [E2].
Grounded generation by Groq. Answers use only retrieved evidence; always verify material claims against the cited source pages.
RETRIEVAL TRACE

Retrieved evidence & source trace

No query yet
Run a query to inspect pgvector similarity, keyword relevance, RRF ranking, source document and PDF page.
SOURCE GOVERNANCE

Controlled corpus, explicit provenance.

No synthetic ESG metrics are seeded into the source database. The ingestion function fetches only the allow-listed official document URLs stored in Supabase.

Provenance contract

Original URLStored for every document
PDF pageStored for every source chunk
SHA-256Recorded for the source files used to establish the corpus
Embedding modelgte-small · 384 dimensions
Content typesource_excerpt only for public retrieval
Fabricated metricsNot permitted in corpus ingestion
RAG EVALUATION

Measure retrieval rather than admiring the chatbot.

The database contains ten expected-source evaluation questions. Run them against the live pgvector index after initialization.

Not run in this browser session.
QuestionExpected source(s)Top retrieved source(s)Hit@5
SYSTEM DESIGN

A database-backed RAG pipeline you can explain.

01Official PDFsSEBI, NSE, company disclosures
→
02Controlled ingestionAllow-listed URLs + page-aware chunking
→
03EmbeddingsSupabase gte-small · 384D
→
04PostgreSQL + pgvectorHNSW vector + GIN full-text indexes
→
05Hybrid RRF retrievalSemantic + keyword fusion
→
06Grounded generationOnly top-k evidence reaches the LLM
VECTOR DATABASE

Real pgvector storage

Document chunks are rows in PostgreSQL and embeddings are stored in a native vector(384) column. HNSW accelerates semantic nearest-neighbour search.

KEYWORD RETRIEVAL

PostgreSQL FTS

A generated tsvector column and GIN index preserve exact ESG terminology such as Scope 3, BRSR Core and assurance.

FUSION

Reciprocal Rank Fusion

Semantic and full-text result lists are merged using RRF so neither exact terminology nor semantic intent dominates by default.

SECURITY

RLS + server-side writes

The browser receives only the public Supabase key. Database writes and retrieval logging happen inside JWT-protected Edge Functions using the service role.

GROUNDING

Evidence contract

The generation endpoint is instructed to use only [E#] evidence, identify boundary/year differences and abstain when retrieval is insufficient.

AUDITABILITY

Retrieval logs

Queries, filters, returned chunk IDs and retrieval latency are recorded server-side for future evaluation and debugging.