Simple text ↔ heuristics & verbs

Integrates collected simple text with the existing _SimpleText similarity toolkit and _Heuristics catalog.

What this is

You collect plain text (verbs, heuristic notes, free-form lines). The system:

1. Loads files under data/simple_text/
2. Extracts verbs (bank + morphology heuristics)
3. Catalogs heuristics (builtin definitions + your notes)
4. Matches queries with SimpleText cosine/Jaccard/Dice/n-gram scores
5. Applies heuristics offline without calling an LLM
6. Optionally applies live via existing _Heuristics::* LLM paths
7. Feeds verb_count / heuristic scores into decision-tree intent features
8. Can import wm_simple_text working-memory rows and export corpus JSONL

File formats

Put files in data/simple_text/:

Samples ship with the repo; replace/extend with your collection.

Commands

simpletext status
simpletext defaults
simpletext load data/simple_text/collected.txt free
simpletext load-dir data/simple_text
simpletext verbs
simpletext heuristics
simpletext extract always verify before you train
simpletext match search memory first then plan
simpletext apply anchoring the first price I saw was $9.99
simpletext features decide and train the model carefully
simpletext wm
simpletext export
simpletext save

heuristic offline auto if everyone is buying it must be good
verbs from plan execute verify integrate

Aliases: stbank …, simple_text ….

Live vs offline heuristics

Offline never requires network; live uses the existing LLM heuristic implementations.

Integration map

Lua

print(ai_simpletext_status())
local verbs = ai_simpletext_verbs("plan execute verify")
print(ai_simpletext_heuristic("auto", "the first offer anchors the negotiation"))

Files

• SimpleTextBank.hpp / .cpp
• data/simple_text/verbs.txt, heuristics.txt, collected.txt
• Existing: SimpleText.*, Heuristics.*