GameCult/AquaSynth
Watch
9
0
Fork
You've already forked AquaSynth
 0
Code Issues Pull requests 4 Projects Releases Packages Wiki Activity Actions

Embed primitive timeline facts in IPA trials

Browse source
This commit is contained in:
Meta Vens 2026-05-31 00:41:28 +01:00
commit f16ea641af
11 changed files with 392 additions and 21 deletions
Download patch file Download diff file Expand all files Collapse all files

28
docs/ipa-vocal-tract-roadmap.md
View file

@ -640,8 +640,9 @@ and stores typed `IpaTrialResult` records in `ipa-trial-results.cc`.
The `.cc` store is per-record CultCache MessagePack data, not a hand-written
JSON ledger. `IpaTrialResult` records include hypothesis text, candidate patch
URI, reference/candidate artifacts, primitive timeline URI, metrics, evaluator
summary, verdict, known contamination, and timing receipts. This is the trial
result database that later hypothesis workers consume.
summary, verdict, known contamination, timing receipts, and extracted primitive
timeline facts. This is the trial result database that later hypothesis workers
consume.
The first five seed trials were run locally on 2026-05-29 under
`artifacts/parity/ipa-trials/20260529T214127955/five-seed-trials`. The first
@ -672,13 +673,14 @@ workers to hand-edit evidence:
against the PT fixtures, and upserts typed `IpaTrialResult` records.
- `search` is the semantic retrieval surface for the `.cc` store. It expands
speech-control vocabulary such as vowel/nasal/fricative/stop, indexes trial
summaries plus referenced local artifact snippets into Qdrant with Ollama
embeddings when the local vector stack is available, ranks records with
vector and metric-aware evidence bias, and writes a compact markdown result
set. If Qdrant or Ollama is down, it falls back to the lexical/metric ranker
and names the fallback in the report.
summaries, embedded primitive timeline facts, and referenced local artifact
snippets into Qdrant with Ollama embeddings when the local vector stack is
available, ranks records with vector and metric-aware evidence bias, and
writes a compact markdown result set. If Qdrant or Ollama is down, it falls
back to the lexical/metric ranker and names the fallback in the report.
- `show` drills into one trial or candidate with full metrics, artifacts,
known contamination, hypothesis, and evaluator summary.
known contamination, hypothesis, extracted primitive timeline facts, and
evaluator summary.
- `dump` remains an audit escape hatch, not the normal agent memory path.
The default local vector organ mirrors VoidBot's retrieval pattern: stable
@ -701,6 +703,16 @@ as stop, nasal, fricative, or vowel is also present. This keeps `mix-p` useful
for plosive seams without letting every mixed candidate impersonate stop
evidence.
Primitive timeline facts are now part of the CultCache record itself, not only
CSV attachments. The extractor records contact closure/opening/reservoir/release
peaks and release block, source pressure/flow peaks, branch admittance/exchanged
flow, radiation boundary/output peaks, path area/delay, path energy, and
passivity ratios. `search`, `show`, `dump`, and trial summary CSVs surface the
release/radiation/passivity witnesses directly. A hypothesis about stops,
fricatives, nasals, vowels, or radiation is not strong evidence unless it can
point to these facts or explicitly say that the required primitive witness is
missing. The machine is allowed to be ugly; it is not allowed to be mysterious.
`tools/run-ipa-trial-loop.ps1` orchestrates the outer loop with external
`codex exec` workers. Each round searches accumulated trial memory, asks a
hypothesis worker to write a batch of candidates, scores those candidates