Every chapter of the Camatkāra Cintāmaṇi is here. We have not established how long the work is, so we cannot say this is all of it. WHICH PRINTING OUR SOURCE IS. The Sanskrit here was machine-parsed from a sanskritdocuments.org e-text, and that e-text is a keyed transcription of ONE identified printing: the comparative-study edition of the Camatkāra Cintāmaṇi with Dharmeśvara Mālavīya's Anvayārthaprabodhapradīpa ṭīkā and Durgāśaraṇa's Maṇiprabhā notes, Motilal Banarsidass, Varanasi 1975. Tier 'etext' licenses the app to say "our source reads…" only when the printing is named, and until 2026-09-08 it was not named because nobody had identified it. It follows that this printing is a CORRECTOR of our keying and NOT an independent witness to the text: a place where it agrees with us proves only that the keyer read it correctly.
WHAT WAS ACTUALLY READ, as of 2026-09-08. Page images of two genuinely independent printings were opened and read against 39 of the 114 verses: the bhāṣāṭīkā edition of Dhanuṣadhārī Miśra (Baijnath Prasad, Benares, Saṃvat 1972 = 1915/16, public domain) and the bhāṣāṭīkā of Rājārāma Tripāṭhī (Navalkishore Press / Tej Kumar Book Depot, Lucknow). Two further printings — Sapkar's Marathi edition of 1865 and a Gujarati-press edition — were opened for chapter 1 only. Thirty-nine verses were corrected as a result and seventeen disagreements are recorded in the chapters' ledgers. All ten chapters now carry a visualValidation record; NONE is audited, because the retrospective visual gate (scripts/translate/workflows/visual-gate.js) has never been run over this text and eye-reading against rendered page images is not that gate. The remaining 75 verses have still never been laid against any printing.
WHAT IS STILL WRONG, said plainly. The metrical hole is closed: every one of the 114 mūla strings now carries exactly 48 akṣaras, which is what bhujaṅgaprayāta requires, and the twelve verses that were still short or long on the morning of 2026-09-08 were each ruled on the ink that afternoon. That is a count, not a scansion — the counter answers "how many vowel nuclei does this string hold?", and it was checked first against six strings whose wrong values were already on record before it ran. The twelve changed verses were additionally weighed by hand, guru by position, and all forty-eight pādas scan. What 48 akṣaras does NOT prove is that every word is the right word: a substitution of equal weight passes the count silently, and one such substitution was caught this same day — at 8.5 an earlier correction had deleted a negation the text carries, and both printings restored it. Two verses were ADDED on 2026-09-08 — a thirteenth Saturn verse (published as 8.10a) and the closing phalaśruti (published as 10.phalaśruti) — which were never in the e-text and which five and four printings respectively carry; those two alone are tier 'constituted'. The work's total length is still not established from any edition: no printing we hold carries a śloka index, a verse-count table or a colophon stating one, so `complete` stays null rather than being computed from a number nobody read.
114 of this text's 114 verses carry an English translation; 10 of 10 chapters are begun or finished (1–10). We have NOT established how long this work is, so we cannot say whether this is all of it. 2 verse(s) are tier 'constituted' (two independent exemplars, an adjudication ledger, a visual gate) and the rest are tier 'etext' (one encoded source, machine-parsed). The two are different claims and the reader is told which.
114 verses in the original Sanskrit with our own English translation, which is made from that Sanskrit and not adapted from any existing translation. Machine-parsed from sanskritdocuments.org e-text (chamatkarachintamani.html) by scripts/corpus/ingest-etext.mjs on 2026-09-01. THE E-TEXT IS A KEYED TRANSCRIPTION OF THE MLBD COMPARATIVE EDITION, VARANASI 1975 (Dharmeśvara Mālavīya's Anvayārthaprabodhapradīpa ṭīkā with Durgāśaraṇa's Maṇiprabhā notes) — established from the two documents themselves: the e-text's dropped front matter names both commentaries, and the 97 Hindi bhāva rubrics reproduced in `notes` are that 1975 editor's headings. That printing is therefore a corrector of our keying and not a witness independent of it. Segmentation rule: Body bounded to the page's <PRE id='content'> block BEFORE any tag stripping — the page carries a second <pre class='inf'> of encoder metadata plus an <h1> that duplicates the title, and a whole-file strip draws the duplicate in. Inside it, tags are stripped, lines trimmed, and lines with no Devanagari dropped (the encoder's Latin credit line, and nothing else). Body starts at the unique maṅgalācaraṇa rubric, which is also chapter 1's opening rubric and so is kept. The six front-matter lines before it are dropped BY NAME (title, two invocations, three lines naming the edition and its two commentaries — neither commentary is in the file), and the drop is asserted: it must contain no ॥ marker, and the first verse terminator must sit exactly two lines past the boundary. That guard is load-bearing — two front-matter lines end in a single daṇḍa and are shaped exactly like a verse's first half, so a walk from line 1 manufactures a phantom verse in chapter 1. Verse boundary = the terminal ॥N॥ marker, anchored at end of line and TRUNCATED at the match so nothing after it can enter the mūla. Every verse is exactly two lines, the first closing on a single daṇḍa; the two halves join on a single space. Runs of whitespace inside a verse collapse to one space — this alters no akṣara and fixes the one double space the source sets, in ch9 (Rāhu) v7 at रुजा धातुपाकादिना. Chapter boundary = the RESET of the verse number to १, giving ten sections. The rubrics are NOT usable as boundaries: Śukra's section has no opening rubric at all, four of the Moon's bhāva rubrics have the same shape as a section rubric, and a rubric-driven split therefore folds Venus's twelve verses into Jupiter's chapter and reports a plausible ten. Chapter titles are this ingest's editorial labels, assigned positionally, because the file's own section rubrics are inconsistent in wording and language and one is missing. They are VERIFIED, not assumed: each chapter's label must be named by a rubric that chapter itself carries, and all ten match their own label and no other. APPARATUS IS NOT SILENTLY DISCARDED: the 100 rubrics are recorded verbatim as `edition` notes on the verse each heads — 94 verses carry one, 3 carry two (ch3 v1, ch8 v1, ch10 v1), 15 carry none, which is 94 + 6 = 100 rubrics over 97 noted verses. They are the printing's, largely in modern Hindi — 'धनभावस्थ शनि के फल', with के/का/में inside a Sanskrit text — and never enter the mūla, because the mūla holds what the source prints as the text and not what it prints about the text. Nothing else is discarded. Every line of the body is classified as a verse half or a rubric, the totals are asserted to account for the file exactly, and a line that matches no rule THROWS rather than being skipped. The source's orthographic defects were originally left unrepaired — 24 ण्ग for ङ्ग, भत्त for भट्ट, धनश्याम° for घनश्याम°, and the two interior daṇḍas at ch2 v3 and ch7 v10 — and that policy was REVERSED on 2026-09-08, because it published Devanāgarī that is not Sanskrit. ण्ग is repaired to ङ्ग on 17 verses (24 tokens); घनश्याम° and घनं are restored at 1.1–1.2 with the e-text's line kept in `diplomatic`; ch7 v10's two interior daṇḍas went with the intruded gloss `। गोत्रबीजम् ।` that caused them, likewise recorded in `diplomatic`. ch2 v3's stray daṇḍa sits inside a pāda that scans continuously and is left as printed. Every repair carries a revisions[] entry on its verse naming what decided it. TWO VERSES WERE ADDED that this e-text does not contain — the thirteenth Saturn verse (8.10a) and the closing phalaśruti (10.phalaśruti) — so the corpus now holds 114 verses against the e-text's 112, and those two are the only records here that did not come through this ingest.
Not yet independently audited. Every reading here was established and recorded by the process described above, but no fresh reader has re-derived these verses from the plates without seeing our record. When that was done for the first text we published, it found real errors — so treat this one as a first edition and check anything you intend to rely on. The evidence behind every verse is committed and readable.
34 reading disagreements were adjudicated and are recorded verse by verse. Every akṣara was legible.
Use it freely — CC BY 4.0. This Sanskrit transcription and this English translation are released under a Creative Commons Attribution 4.0 licence. Copy them, print them, build on them; the only condition is attribution to Karma Sakha (karmasakha.com). The underlying work is ancient and in the public domain everywhere.
If you find an error, it is ours and we want it. Every reading on this page is recorded with the evidence behind it, so a disagreement can be argued about concretely rather than asserted.