VulcanRAG/Legal Research
— judgments — citation edges
PRECEDENT-AWARE LEGAL RESEARCH

Read the judgment.
Trace what it relies on.

Search Telangana High Court decisions with preserved court reasoning, party arguments, conclusions, and citation-treatment signals.

Try

Your evidence set will appear here

Every result remains linked to its local corpus record and stable document ID.

CORPUS BROWSER

Judgments

Search complete text and filter the local record set.

Loading corpus…
PRECEDENT NETWORK

Citation Graph

Explore outbound authorities and later judgments that cite a decision.

⌘
SELECTED JUDGMENT

No judgment selected

Enter a corpus document ID or search by title.

Selected In corpus External precedent
MODEL SERVICE

DGX Connection

Verify that this application can reach the private model server before configuration or ingestion.

○
LIVE CONNECTION

Not checked

Run a connection check to query the configured private Ollama endpoint.

SERVICE

Connection details

APPROVED INVENTORY

DGX model availability

The connection check lists only models approved for this application and never starts ingestion.

SYSTEM SETUP

Configuration

Connect model serving, assign roles, and choose ingestion behaviour before starting the corpus job.

DGX SPARK

Model server connection

Not configured
Enter the remote model API address.
For authenticated endpoints, set VULCAN_INGEST_API_KEY before starting the app. Credentials are never saved here.
DGX Spark model rolesEmbeddings only: bge-m3:latest (1,024 dimensions). Reserved general synthesis: qwen2.5:32b-instruct. Highest-quality/heaviest synthesis: llama3.3-abliterated:70b-16k. Lighter options: gemma4:e4b and gemma3:4b.
INGESTION

Chunking and batches

ANSWER ENGINES

Synthesis availability

SAVE CONFIGURATION

Validate the model connection before ingestion

Saving does not start or modify the corpus.

DATA PIPELINE

Ingestion

Track source indexing, semantic-vector readiness, and model-serving requirements.

…
SEMANTIC INGESTION

Checking local pipeline…

Reading the judgment index, available models, and storage capacity.

1Complete

Source corpus

Parsed JSON, Markdown, and raw HTML remain the immutable source record.

2Complete

Search and citations

Full text and authoritative scraper-provided citation treatments are indexed locally.

3Pending

Semantic vectors

Judgment sections must be embedded before semantic and mixed retrieval is available.

4Required

Embedding and synthesis models

The embedding model builds and queries semantic vectors; selected chat models synthesise answers without changing the authoritative citation graph.

NEXT ACTION

Semantic ingestion is not configured

Choose and install the local embedding and extraction models before starting a resumable batch.

Only the configured judgment corpus is eligible for this pipeline; unrelated local archives remain outside its scope.

LOCAL INDEX

Corpus Status

Coverage, citation treatments, benches, and index provenance.

● SQLite integrity verified
COVERAGE

Judgments by year

SIGNALS

Citation treatments

NETWORK CENTRALITY

Most cited document IDs

BENCHES

Highest record counts

ARCHITECTURE

What is indexed