Start with a question
Find an author’s papers, ask about a calendrical scheme, or locate a Sanskrit expression.
A guide to the corpus and its connections
Patra Darpaṇa MCP Explorer
Ask. Find. Follow the evidence.
One conversation.
Several ways into the texts.
Patra Darpaṇa MCP lets a connected assistant explore Sañcaya’s Indic texts and research papers through complementary retrieval tools.
CatalogFind papers and their metadata
LexicalFind words, patterns and matching files
SemanticFind passages similar in meaning
EntitiesFollow names, mentions and curated connections
Follow useful connections
Resolve names and aliases. Explore the ontology, then search the corpus for supporting passages.
Return to the source
Ask for readable links and locations. A curated relationship or similar passage is a starting point, not proof.
This page is an informational, read-only explorer. It does not connect to your assistant, perform an OAuth login, or report live MCP health.
Names, ideas and their connections
Curated ontology assertions. Arrows read from source to destination; they are not relationships extracted from corpus occurrences.
Entity neighbourhood
Drag the background to pan. Use the zoom buttons or Ctrl + scroll. Select a node for details.
Visible relationships
| Source | Relationship → | Destination |
|---|
Bring the corpus into your conversation
Use an MCP-compatible client that supports remote HTTP transport and OAuth.
Preferred OAuth endpoint
Connect your assistant
https://sanchaya.rasowshi.us/mcp- In your client’s MCP or connector settings, add the URL above.
- Choose OAuth and sign in with Google when prompted.
- Your Google email must be on the service’s allowlist. Ask the service operator for access.
- Ask your assistant to use Patra Darpan MCP explicitly.
A productive first conversation
Ask for evidence, not just identifiers
Tell your assistant to read the search guide and inspect coverage before searching. Ask it to cite work or paper titles with hyperlinks and verse, section or line locations.
For metadata, use author and catalog tools first. For occurrences, use lexical search and paging. For meaning, try semantic search and verify its passages.
Try a conversation
Using Patra Darpan MCP, search lexically for file:Jyo type:filename सूर्य. Return the filenames of matching files without content snippets.
Using Patra Darpan MCP, look up Dhaniṣṭhā. Show its ontology attributes and labelled relationships. Keep curated assertions separate from corpus evidence.
Using Patra Darpan MCP, find passages about Śraviṣṭhā and Dhaniṣṭhā in texts and papers. Try explicit script and spelling variants, follow continuation when available, and cite sources. Explain the limits of the search.
Different indexes, different coverage
Published configuration summary · 5 October 2026. This is not a live release inventory.
| Layer | What it covers | Best used for |
|---|---|---|
| Catalog · SQLite | Full assembled paper metadata; text records in the release catalog. Having metadata does not imply indexed content. | Authors, titles, years and source URLs |
| Lexical · Zoekt | Configured Sañcaya files, including exported paper Markdown. Wider scope than the passage and entity projections. | Words, regex, counts, continuation and filenames |
| Semantic · E5 / Qdrant | Release-selected paper and Sanskrit text chunks, including Jyotisham and selected anchor texts. Not all Sañcaya. | Passages similar to a natural-language question |
| Entity projection | Explicit ontology aliases matched within release chunks; a registry and paged mentions. | Name resolution and known occurrences |
| Ontology · Jyotiṣa | Curated entities, attributes and canonical relationships. The versioned snapshot on this page is separate from live release data. | Connections to explore and assertions to verify |
Check the active release
Ask your connected assistant to call get_corpus_info. Use its actual scope, revision and counts rather than treating this page as the deployed inventory.
Zero does not mean absent
Missing aliases, case endings, OCR and partial chunk coverage can yield zero entity mentions even when lexical search finds relevant text.
Similarity is not proof
E5’s Sanskrit quality was assessed in a small pilot. Semantic results are ranked suggestions, not a complete occurrence inventory or historical verification.
Explore Sanskrit carefully
Try script equivalents, plausible declensions and joined or separated sandhi forms. Treat generated forms and compound components as hypotheses; verify their sense in passages.
Fetch only where passages are available, follow paging, and check response limits before claiming completeness. Ontology provenance is often entity-level curation, not a citation for every edge.
Read the search guide ↗