Provider Extraction

Promemo can extract structured memory from messy notes, session summaries, handoffs, or snapshots using a configured OpenAI-compatible provider.

Configure The Provider

Provider settings live in:

.promemo/config.toml

Example:

[project]
name = "Promemo Project"

[provider]
kind = "openai-compatible"
base_url = "https://api.openai.com/v1"
model = "gpt-4.1-mini"
api_key_env = "OPENAI_API_KEY"
timeout_seconds = 60

Set the configured API key environment variable:

export OPENAI_API_KEY="..."

Extract From Notes

Preview extracted memory without writing files:

cat <<'EOF' | promemo extract authentication --stdin --dry-run
# Authentication notes
- Refresh tokens rotate on every use.
EOF

Write extracted memory:

cat <<'EOF' | promemo extract authentication --stdin
# Authentication notes
- Refresh tokens rotate on every use.
EOF

Read source text from stdin:

printf '# Authentication notes\n- Refresh tokens rotate on every use.\n' | promemo extract authentication --stdin --dry-run

What Extraction Saves

Provider extraction:

1. Reads unstructured text.
2. Sends it to the configured provider with the MemoryInput schema.
3. Validates the provider response.
4. Uses the same preview/save pipeline as save-json.
5. Does not store raw transcript text.