Gochara Source Crosswalk
Verified layer
The seven classical Grahas' favourable transit houses are verified against Varahamihira's Brihat Samhita, Chapter 104, stanza 4, in the N. Chidambaram Iyer translation (1884). The stanza lists each Graha's favourable houses from Janma Chandra and concludes that every Graha is favourable in the 11th.
| Graha | Implemented favourable houses | How stanza 4 supports the set |
|---|---|---|
| Surya | 3, 6, 10, 11 | 3, 6 and 10 listed; 11 supplied by the concluding rule |
| Chandra | 1, 3, 6, 7, 10, 11 | 1, 3, 6, 7 and 10 listed; 11 supplied by the concluding rule |
| Mangala | 3, 6, 11 | 3 and 6 listed; 11 supplied by the concluding rule |
| Budha | 2, 4, 6, 8, 10, 11 | 2, 4, 6, 8 and 10 listed; 11 supplied by the concluding rule |
| Guru | 2, 5, 7, 9, 11 | 2, 5, 7 and 9 listed; 11 supplied by the concluding rule |
| Shukra | 1, 2, 3, 4, 5, 8, 9, 11, 12 | 6, 7 and 10 called malefic; 11 supplied by the concluding rule |
| Shani | 3, 6, 11 | 3 and 6 listed; 11 supplied by the concluding rule |
The ledger claim is gochara.favourable_houses. This verifies the table, not every interpretive sentence produced from it.
Verified Vedha layer
Mantreswara's Phaladeepika, Adhyaya XXVI, slokas 3–8, states the favourable houses and their corresponding Vedha houses planet by planet. The same passage explicitly exempts Shani from obstructing Surya, Surya from obstructing Shani, Budha from obstructing Chandra, and Chandra from obstructing Budha. The consolidated table and father–son explanation also appear in Jataka Parijata, Volume III, Adhyaya XIII, note to sloka 60, printed pages 833–834.
| Graha | Implemented favourable → Vedha pairs | Phaladeepika locator |
|---|---|---|
| Surya | 3→9, 6→12, 10→4, 11→5 | XXVI.3 |
| Chandra | 1→5, 3→9, 6→12, 7→2, 10→4, 11→8 | XXVI.4 |
| Kuja | 3→12, 6→9, 11→5 | XXVI.5 |
| Budha | 2→5, 4→3, 6→9, 8→1, 10→8, 11→12 | XXVI.6 |
| Guru | 2→12, 5→4, 7→3, 9→10, 11→8 | XXVI.7 |
| Shukra | 1→8, 2→7, 3→1, 4→10, 5→9, 8→5, 9→11, 11→6, 12→3 | XXVI.8 |
| Shani | 3→12, 6→9, 11→5 | XXVI.5 |
The ledger claim is gochara.vedha_tables. This verification stops at the seven classical Grahas; it does not silently authorize the project's node policy.
Verified node houses
Phaladeepika XXVI.2 concludes that Rahu and Ketu are to be treated like Surya. The implementation therefore uses houses 3, 6, 10 and 11 for both nodes. The gochara.nodes claim is verified for this favourable-house layer.
This locator does not say whether Rahu or Ketu causes or receives Vedha. The current no-node-Vedha behavior remains an unverified conservative sub-rule and does not inherit the authority of XXVI.2.
Partially verified named-Shani layer
| Layer | Claim | State | Boundary |
|---|---|---|---|
| Named Shani conditions | gochara.named_shani_conditions | partially_verified | Phaladeepika XXVI.1 establishes Janma Chandra as the reference; XXVI.22–23 describes adverse effects in houses 1, 2, 4, 8 and 12. It does not supply the configured grouping, names, phase labels or advice. |
This distinction corrects a former frontend error: when users supplied both a Moon sign and lagna, the browser generated named Shani flags independently from both. Named conditions are now calculated only from the Moon sign. Lagna remains available as a secondary lens for ordinary house-by-house gochara verdicts.
The terms “Sade Sati,” “Ashtama Shani,” and “Ardhastama Shani” remain useful conventional shorthand, but the product does not call them verbatim classical labels. Likewise, “rising,” “peak,” and “setting” are presentation labels rather than claims found in the cited slokas.
Generated interpretation boundary
The website may publish a daily LLM-written interpretation above the computed fallback reading. Its transits_cited graha, house and verdict triples are checked against the engine. That check does not prove every prose sentence or piece of advice. Those fields carry the separate daily_horoscope.llm_interpretation heuristic claim and must not inherit the textual authority of the transit table.
The 2026-07-21 published artifact demonstrated the boundary failure that this contract fixes: while its declared transit triples were accepted, uncited prose introduced a late-afternoon contract window, a 48-hour decision delay, and claims about investment returns, health and intraday clarity. None of those facts existed in the generator input.
Generation now fails closed on duplicate or missing rashis, empty citations, cited Grahas absent from the prose, precise intraday/waiting-period claims, and selected medical, legal, investment or contract directives. These controls reduce unsupported output; they are not represented as semantic proof.
The MCP response publishes all four claim IDs and explains the split. Product copy must not describe the entire verdict system as “the Brihat Samhita table,” or imply that verified Vedha evidence also settles the two open layers.
Engineering consequence
Exact table tests bind the classical source crosswalk to the implementation, while each adjacent conventional layer retains its own status and can be promoted without rewriting the calculation contract.