@polycode-projects/the-mechanical-code-talker 1.9.1 → 1.9.2
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +0 -15
- package/corpus/seon/README.md +1 -2
- package/package.json +1 -1
- package/src/chat.mjs +1 -3
package/README.md
CHANGED
|
@@ -392,21 +392,6 @@ once automatically after seeding and on demand; the entailed facts are
|
|
|
392
392
|
**low-trust and retractable** (never outranking a stated fact) and this never runs
|
|
393
393
|
on the chat's hot path.
|
|
394
394
|
|
|
395
|
-
## What tmct deliberately is NOT
|
|
396
|
-
|
|
397
|
-
- **It is not an indexer.** tmct keeps no codebase index of its own. It
|
|
398
|
-
consumes a graph via a provider seam (`fetchEntities` and friends) — building
|
|
399
|
-
that graph is a different tool's job. tmct's job is the *conversation*.
|
|
400
|
-
- **It is not a reasoning model.** Where it "reasons", it does so by
|
|
401
|
-
*calculation* surfaced as prose ("there are a lot of tests for a codebase of
|
|
402
|
-
that size"). It is deterministic, explainable, and cheap. Even its forward-chaining
|
|
403
|
-
entailment (`tmct syllogise`) is mechanical OWL rule materialization applied
|
|
404
|
-
offline, rule-by-rule and retractable, not an LLM. There is **no LLM anywhere
|
|
405
|
-
in the product**. (An LLM-as-judge exists only in the offline eval harness
|
|
406
|
-
that tunes tmct, see `SKILL_BENCHMARK_CEFR_ENGLISH.md`, never in the product path.)
|
|
407
|
-
- **It never guesses silently.** When it cannot resolve your question it says
|
|
408
|
-
so and nudges you toward a query it *can* answer.
|
|
409
|
-
|
|
410
395
|
## Install & use
|
|
411
396
|
|
|
412
397
|
```bash
|
package/corpus/seon/README.md
CHANGED
|
@@ -111,8 +111,7 @@ This directory is **data only**; wiring it into seeding is the coordinator's
|
|
|
111
111
|
the seon tier the ordering only affects the ConceptNet tail.
|
|
112
112
|
3. **Provenance caveat:** `toFacts` currently hard-codes the provenance string
|
|
113
113
|
`corpus:conceptnet <rel>`. If the coordinator wants seon facts tagged as their own
|
|
114
|
-
source (`corpus:seon`), that is a one-line parametrisation of `toFacts`
|
|
115
|
-
`seedMemory` — out of scope for this data-only directory.
|
|
114
|
+
source (`corpus:seon`), that is a one-line parametrisation of `toFacts`
|
|
116
115
|
|
|
117
116
|
## Regenerating / extending
|
|
118
117
|
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@polycode-projects/the-mechanical-code-talker",
|
|
3
|
-
"version": "1.9.
|
|
3
|
+
"version": "1.9.2",
|
|
4
4
|
"private": false,
|
|
5
5
|
"type": "module",
|
|
6
6
|
"description": "The Mechanical Code Talker (tmct) — a tolerant, offline, $0 chat surface that guides you toward precision queries about a software repository. ELIZA/PARRY-style but domain-obsessed with code. No model calls; no codebase index of its own.",
|
package/src/chat.mjs
CHANGED
|
@@ -2377,9 +2377,7 @@ async function unknownObjectFallback(payload, { memoryDir, sessionId, lexicon },
|
|
|
2377
2377
|
* Sits strictly UPSTREAM of the pre-existing TEACH_PROPERTY_RE gap (the
|
|
2378
2378
|
* wrapped-only surface that mints ANY bare complement word with zero
|
|
2379
2379
|
* grounding check at all, e.g. "remember that zorp is florpy" —
|
|
2380
|
-
* Verification finding 3, PLAN_TAUGHT_RELATIONS.md):
|
|
2381
|
-
* close that gap (out of scope, a deliberate separate operator decision),
|
|
2382
|
-
* only adds a properly-grounded alternative ahead of it.
|
|
2380
|
+
* Verification finding 3, PLAN_TAUGHT_RELATIONS.md): .
|
|
2383
2381
|
*
|
|
2384
2382
|
* IMPLEMENTATION ADJUSTMENT found live (not in the original plan text): a
|
|
2385
2383
|
* bare "module is banana" (a KNOWN lexicon-noun subject, NO article, NO
|