Commit graph

12 commits

Author SHA1 Message Date
jlightner
ca5aa3dec0 feat: Rewrote stage5_synthesis.txt with v2 body_sections (list-of-objec…
- "prompts/stage5_synthesis.txt"
- "prompts/stage5_synthesis.20260403_005044.bak"

GSD-Task: S01/T02
2026-04-03 00:52:48 +00:00
jlightner
06b8bdd6ac prompt: stage5 synthesis v4 — instructive voice, name discipline, merge thresholds
- Rewrote voice from third-person narrative ("Keota does X") to instructive
  ("Route the effect at 100% wet"). Body prose now reads like a lesson book.
- Hard rule: creator name appears in title/summary only, max once in body
  (for quote attribution). Fixed JSON example that modeled heavy name usage.
- Added orientation-first section rhythm: brief definition before diving into
  method, prevents run-on feel.
- Page minimum thresholds: 3+ sections, 400+ words, 3+ moments. Prevents
  stub pages from thin categories.
- Strengthened merge guidance: prefer fewer rich pages over many stubs.
- Updated all examples to model instructive phrasing.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-02 22:14:43 +00:00
jlightner
84e617ab64 stage5 prompt: make section ordering a hard constraint with explicit wrong/correct examples
Validation/quality-check sections can NEVER precede construction sections.
Added concrete wrong vs correct ordering example using the exact snare design
case that failed. Elevated from 'typically' guidance to non-negotiable rule.
2026-04-01 11:35:59 +00:00
jlightner
78c01b22ca stage5 prompt: reduce creator name repetition — use pronouns after establishing attribution 2026-04-01 11:25:45 +00:00
jlightner
9e1feae76d stage5 prompt: add explicit section ordering guidance — follow the workflow
Sections should mirror the actual production workflow: foundations before
finishing, construction before glue, sound sources before processing before
mix-bus treatment. Includes the test: 'would a producer follow these steps
in this sequence?' and a natural flow template (framework → construction →
combining/refining → quality checks).
2026-04-01 11:23:20 +00:00
jlightner
2b06828aaa stage5: replace synthesis prompt with v016 (masterclass-recap) + add 100 variant prompts
New prompt combines: embedded documentarian role, distilled-knowledge framing,
conversational authority voice, problem-solution section structure,
context-wrapped specifics, problem-driven teaching rhythm, any-skill-level
reader model, insight-first summary, and engagement emphasis.

100 variant prompts generated across 9 dimensions of variation for future
A/B testing. Generator script included for reproducibility.
2026-04-01 10:49:16 +00:00
jlightner
d75ec80c98 optimize: Stage 5 synthesis prompt — round 0 winner (0.95→1.0 composite)
Applied first optimization result: tighter voice preservation instructions,
improved section flow guidance, trimmed redundant metadata instructions.
13382→11123 chars (-17%).
2026-04-01 10:15:24 +00:00
jlightner
c344b8c670 fix: Moment-to-page linking via moment_indices in stage 5 synthesis
When the LLM splits a category group into multiple technique pages,
moments were blanket-linked to the last page in the loop, leaving all
other pages as orphans with 0 key moments (48 out of 204 pages affected).

Added moment_indices field to SynthesizedPage schema and synthesis prompt
so the LLM explicitly declares which input moments each page covers.
Stage 5 now uses these indices for targeted linking instead of the broken
blanket approach. Tags are also computed per-page from linked moments
only, fixing cross-contamination (e.g. "stereo imaging" tag appearing
on gain staging pages).

Deleted 48 orphan technique pages from the database.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-04-01 00:34:37 -05:00
jlightner
4b0914b12b fix: restore complete project tree from ub01 canonical state
Auto-mode commit 7aa33cd accidentally deleted 78 files (14,814 lines) during M005
execution. Subsequent commits rebuilt some frontend files but backend/, alembic/,
tests/, whisper/, docker configs, and prompts were never restored in this repo.

This commit restores the full project tree by syncing from ub01's working directory,
which has all M001-M007 features running in production containers.

Restored: backend/ (config, models, routers, database, redis, search_service, worker),
alembic/ (6 migrations), docker/ (Dockerfiles, nginx, compose), prompts/ (4 stages),
tests/, whisper/, README.md, .env.example, chrysopedia-spec.md
2026-03-31 02:10:41 +00:00
jlightner
7aa33cd17f fix: Fixed syntax errors in pipeline event instrumentation — _emit_even…
- "backend/pipeline/stages.py"

GSD-Task: S01/T01
2026-03-30 08:27:53 +00:00
jlightner
f99ac1b8b9 prompts: Rewrite all four pipeline stage prompts for quality and domain awareness
- Stage 2: Add domain context, granularity guidance, unstructured content handling
- Stage 3: Add extract/skip framework, summary quality standards, fewer-richer directive
- Stage 4: Add production-session classification principles, ambiguity resolution examples
- Stage 5: Add voice/tone guidance, anti-generic section names, signal chain detail, anti-filler rules

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-30 07:27:35 +00:00
jlightner
b5635a09db feat: Created 4 prompt templates and implemented 5 Celery tasks (stages…
- "prompts/stage2_segmentation.txt"
- "prompts/stage3_extraction.txt"
- "prompts/stage4_classification.txt"
- "prompts/stage5_synthesis.txt"
- "backend/pipeline/stages.py"
- "backend/requirements.txt"

GSD-Task: S03/T02
2026-03-29 22:36:06 +00:00