ROOTED IN SOURCES. OPEN TO POSSIBILITY.
A growing foundation for exploring Tamil Bhakti, Sanskrit scripture, and the teachings that connect them.
Pramāṇa brings texts, context, and evidence together—so new experiences can grow without losing sight of where their knowledge comes from.
THE IDEA IN ONE SENTENCE
The long-term goal is a reusable, auditable foundation for readers, researchers, curators, and builders—not a single app or a single way of understanding a tradition.
01 / THE CORPUS TODAY
Pramāṇa 1.0 brings nine components together at their declared maturity. These are different kinds of evidence—not one interchangeable total.
The Shaiva textual foundation, including Tevaram, Tiruvacakam, Tirukkovaiyar, and Periya Puranam.
68 episodes and 1,763 evidence edges, with literary, edition, and epigraphic authority distinguished.
Deivathin Kural, Volumes 1–7. Curated summaries and source references, not a redistribution of the source text.
| Collection | Coverage | What that means |
|---|---|---|
| Tirumurai 1–7 · Tevaram | Source-preserved beta | Project Madurai source family with IFP/PIFI comparison. The 8,240 running-number convention is retained; it is not a certified count of unique texts. |
| Tirumurai 8 | 661 + 400 source units | Tiruvacakam and Tirukkovaiyar, with source order and printed-number anomalies preserved. |
| Tirumurai 9–11 | 301 / 3,046 / 1,419 | Source units or loci, depending on the edition. Tirumurai 10 explicitly retains the missing locus 851 and other edition anomalies. |
| Tirumurai 12 · Periya Puranam | 4,286 structural units | 13 carukkams and 68 story units. A narrative in a preserved text is not automatically independently verified history. |
| Andal | 173 pasurams | Tiruppavai 30 + Nachiyar Tirumozhi 143. Source-preserved beta; metadata ready, source-text publication separately gated. |
| Isha Upanishad | 18 Kāṇva mantras | 15 conservative normalized matches and three explicit witness differences at mantras 1, 13, and 14. Witness independence is not assumed. |
| Bhagavatam · Narasimha | 3 source loci | 7.8.29, 7.8.45, and 7.8.49: a second-witness-aligned evidence seed. The full Bhagavatam is outside the 1.0 hardened scope. |
| Deivathin Kural | Volumes 1–7 | 3,329 records attested in Deivathin Kural and 39 supported by earlier witnesses. None are labeled primary-source-verified. Metadata publication approval remains separate. |
Snapshot: GitHub main, 24 September 2026 · commit 6ecad2e · Inspect the release evidence (JSON)
Coverage is not a blanket publication license. Some components support metadata access or internal use while their source text remains subject to separate rights review.
02 / WHY THE FOUNDATION MATTERS
A corpus is an organized collection of material. Pramāṇa adds the context needed to decide what that material can—and cannot—support.
Keep an addressable passage in a named source or edition. Preserve original numbering and record anomalies instead of silently fixing them.
Record comparisons, witness relationships, review decisions, and uncertainty. Agreement between two websites does not by itself establish independence.
Link teachings, people, hymns, episodes, and places to their evidence. Keep traditions, textual variants, and kinds of authority distinct.
Carry citations and evidence status into the experience. A generated explanation should remain distinguishable from a quotation or a curator’s summary.
A CONCRETE EXAMPLE
A Nayanmar trail may connect a saint’s story, a temple, and a hymn. Those connections do not all mean the same thing.
Evidence for what the work narrates. It does not independently prove that the event happened as described.
An edition’s metadata may associate a hymn with a site. That association should keep its source and should not be silently upgraded into an archaeological claim.
Explicitly sourced epigraphic evidence can support a particular historical statement. Its authority applies to that statement, not every story connected to the same place.
The value is in these distinctions surviving every search, visualization, and retelling.
03 / ONE FOUNDATION, MANY EXPERIENCES
Explore existing sites that illustrate different ways to encounter this material. A link here does not imply that every site already runs on Pramāṇa.
Ask Mahaperiyava connects questions with teachings from Deivathin Kural. Pramāṇa’s seven-volume metadata and grounded retrieval work form the project’s teaching-discovery foundation.
Existing Pramāṇa experienceA site about Nayanmar stories, sthalams, and Tevaram. Pramāṇa’s 63-saint evidence graph provides a foundation for source-linked trails and different views of these relationships.
Existing site · related corpus capabilitySandhya Katha offers source-checked Hindu stories for children. A future Pramāṇa integration could help trace a retelling to its passages and distinguish the source from the storyteller’s adaptation.
Existing site · potential integrationVamsha explores Indian family trees. Pramāṇa could eventually supply source-linked relationships, keep differing genealogical traditions visible, and show where the evidence is incomplete.
Existing site · potential integrationThese are examples of interfaces, not a promise that the present corpus supports every story, person, or question on these sites.
04 / PROGRESS, WITHOUT THE FINE PRINT
The latest recorded milestone is Pramāṇa 1.0 RC1. Component readiness, verification depth, and permission to publish remain separate decisions.
IN PLACE
Nine release components, the twelve-Tirumurai textual foundation, the Nayanmar graph, Andal and Isha betas, a bounded Bhagavatam seed, and seven-volume teaching metadata.
The teaching-discovery backend and mobile web integration have a recorded readiness checkpoint.
BEFORE THE FINAL RELEASE
The release contract requires a clean full automated corpus check on the final revision, followed by a check that release artifacts have not drifted, before the final tag.
The September 24 checkpoint records a CI runner incident. This is a dated project snapshot, not a live service-status report.
BEYOND 1.0
Expand independent witness review, resolve publication terms, and strengthen evidence depth. Later scope includes the rest of Divya Prabandham, fuller Bhagavatam coverage, more Upanishads, and Bhagavad Gita hardening.
Future scope is a direction, not a delivery promise. No new major corpus is part of the current release freeze.
A source unit, graph edge, and teaching summary measure different work. They are not added together. “Source-preserved” means traceable to a declared edition, not independently verified history. “Metadata ready” does not mean the underlying text is free to redistribute.
Yes. The teaching app’s recorded readiness evaluation reports 95% top-five recall over 40 supported questions, with separate unsupported and adversarial checks. This is an internal, bounded evaluation, not a claim of 95% accuracy on every question or superiority to other products. The corpus release also has component-specific contract and publication checks.
Reviewed against GitHub main 6ecad2e on 26 September 2026. Read the machine-readable snapshot.
05 / FAIR QUESTIONS
The project should be judged by inspectable evidence and useful experiences—not by the size of its claims.
A database is part of it. The intended value is the curation around it: source-specific identifiers, edition context, witness comparisons, qualified relationships, rights information, and review decisions. If those layers do not make an answer easier to inspect or a new application easier to build, the distinction has not earned its keep.
You can. Pramāṇa addresses a different need: knowing which source supports a statement and where its limits lie. A language model can help find or explain material, but fluency is not evidence. The corpus and its provenance should remain useful when the model changes.
No. Pramāṇa uses explicit authority levels. Much of the 1.0 corpus is a source-preserved beta. A comparison witness may share an editorial lineage with the primary source. Historical truth, source attestation, textual agreement, and doctrinal interpretation are separate questions.
No single harmonized account should replace the evidence. Recensions, variants, and tradition-specific interpretations need to remain identifiable. A disagreement is information to preserve, not a gap for an AI to smooth over.
No blanket claim is made. Code access, curator-metadata terms, and source-text rights are separate. The release supports rights-aware packaging, and several components remain metadata-only or separately gated. A work’s age does not determine the reuse terms of every modern edition or transcription.
No. The 3,368 Deivathin Kural records are curated teaching metadata. They do not redistribute the underlying source text and must not be presented as verbatim speech. Earlier-witness support is recorded only where the review supports it.
No. “Not found in this corpus” describes a coverage limit. It does not prove that a passage, tradition, or event does not exist elsewhere. A responsible interface should make that distinction visible.
06 / BUILD ON WHAT IS HERE
A reader, a map, a comparison tool, a classroom experience. The opportunity is to make the material more accessible while keeping its context intact.
Choose a bounded collection and an audience. Check the evidence status and rights of the records you need. Design the source view alongside the main experience, so readers can follow an answer back to its origin.
Pramāṇa’s structured records, source references, and evidence labels can support retrieval-augmented generation (RAG), reading interfaces, story collections, and knowledge explorers. Each application must respect the evidence status and reuse rights of the material it uses.
Explore a multilingual reading interface, a source comparison view, a saint–temple–hymn map, or a teaching-discovery experience. Use stable record IDs, show source attribution, label adaptations, and provide an honest “not enough evidence” state.
Help identify genuinely independent witnesses, review source-numbering anomalies, distinguish literary and historical evidence, or clarify exact-source reuse terms. Traceable corrections are more useful than an unqualified increase in record count.
Repository on GitHub · access required ↗
The repository is currently private. Collaborators with access can propose a focused change there. Public contributions will open when repository access and contribution terms are established.
A COMMON WAY HOME
For an experience that actually uses the corpus, this small attribution gives readers a path back to the evidence standards and coverage behind it.
Powered by Pramāṇa ↗Copy a link to pramanacorpus.in for your integrated site. Use it after integration; inspiration alone is not a “powered by” relationship.
Each independently hosted site needs to add this link once. It does not appear on other sites automatically.