alive-mine

Mine for Context

Deep extraction from source material. Creates reference bundles, builds extraction plans, tracks what's been mined.

What Gets Mined

• Session transcripts (JSONL files)
• Meeting transcripts (Fathom, Otter)
• Email threads
• Documents, PDFs
• Slack/chat exports
• Any substantial text source

Extraction Targets

For each source, extract:

1. Decisions -- what was decided, by whom, with rationale
2. Tasks -- action items with owners and deadlines
3. People context -- new information about known people, or new people
4. Insights -- domain knowledge that persists across sessions
5. Quotes -- sharp, memorable statements worth preserving
6. Connections -- references to other walnuts, bundles, or projects

Process

1. Read the source. Full file, no skimming.
2. Extract per category. Group findings by type.
3. Present for approval. Nothing routes without the user seeing it.

Extracted from [source]:

Decisions (3):
  1. [decision] -> [[walnut]]
  2. [decision] -> [[walnut]]
  3. [decision] -> [[walnut]]

Tasks (2):
  4. [task] -> [[walnut]]
  5. [task] -> [[walnut]]

People (1):
  6. [person context] -> [[person]]

Insights (1):
  7. [insight] -> [[walnut]] insights

Approve all, or pick numbers to route?

1. Route approved items. Decisions -> log entries. Tasks -> tasks.py. People -> person walnuts. Insights -> insights.md.
2. Mark source as mined. Update bundle manifest mining: field.

Mining State

Track in bundle manifest:

• mining: active -- still gathering sources
• mining: paused -- temporarily stopped
• mining: exhausted -- all known sources captured

Batch Mining

For multiple sources (e.g., scanning all unsaved sessions):

# Find unmined squirrel entries
find "$ALIVE_WORLD_ROOT/.alive/_squirrels" -name "*.yaml" -newer "$ALIVE_WORLD_ROOT/.alive/_squirrels/.last-mined" 2>/dev/null

Process each, present aggregated results.