avocet/config/label_tool.yaml.example

# config/label_tool.yaml — Multi-account IMAP config for the email label tool
# Copy to config/label_tool.yaml and fill in your credentials.
# This file is gitignored.

accounts:
  - name: "Gmail"
    host: "imap.gmail.com"
    port: 993
    username: "you@gmail.com"
    password: "your-app-password"   # Use an App Password, not your login password
    folder: "INBOX"
    days_back: 90

  - name: "Outlook"
    host: "outlook.office365.com"
    port: 993
    username: "you@outlook.com"
    password: "your-app-password"
    folder: "INBOX"
    days_back: 90

# Optional: limit emails fetched per account per run (0 = unlimited)
max_per_account: 500

# cf-orch SFT candidate import — path to the bench_results/ directory
# produced by circuitforge-orch's benchmark harness.
sft:
  bench_results_dir: /path/to/circuitforge-orch/scripts/bench_results

# cf-orch integration — LLM benchmark harness via cf-orch coordinator.
# All keys here override the corresponding environment variables.
# Omit any key to fall back to the env var (see .env.example).
cforch:
  # Path to cf-orch's benchmark.py script
  bench_script: /path/to/circuitforge-orch/scripts/benchmark.py
  # Task and model definition files (yaml)
  bench_tasks:  /path/to/circuitforge-orch/scripts/bench_tasks.yaml
  bench_models: /path/to/circuitforge-orch/scripts/bench_models.yaml
  # Where benchmark results are written (also used for SFT candidate discovery)
  results_dir:  /path/to/circuitforge-orch/scripts/bench_results
  # Python interpreter with cf-orch installed
  python_bin:   /devl/miniconda3/envs/cf/bin/python

  # Connection config — override env vars CF_ORCH_URL / CF_LICENSE_KEY / OLLAMA_HOST / CF_JUDGE_URL / HF_TOKEN
  # coordinator_url: http://localhost:7700
  # license_key:     CFG-AVCT-xxxx-xxxx-xxxx
  # ollama_url:      http://localhost:11434
  # ollama_model:    llama3.2:3b
  # embed_model:     nomic-embed-text   # Ollama embedding model for EmbeddingKNNAdapter
  # judge_url:       http://10.1.10.158:8008  # Sif cf-text — LLM-as-judge secondary scorer
  # judge_url:       http://10.1.10.71:8008   # Heimdall cf-text (alternative)
  # Or set CF_JUDGE_URL. Populates the Judge URL field in the LLM Eval UI automatically.
  # hf_token:        hf_xxxxxxxxxxxxxxxxxxxx  # HuggingFace token — required for gated/terms-restricted models

  # Directory containing per-node profile YAMLs (cf-orch node profiles).
  # Default: derived from bench_script location (../../profiles/nodes).
  # profiles_dir: /Library/Development/CircuitForge/circuitforge-orch/circuitforge_orch/profiles/nodes

# Imitate tab — pull real samples from sibling CF product APIs and run them
# through local LLMs to build a corrections dataset.
# ollama_url defaults to cforch.ollama_url if omitted here.
imitate:
  ollama_url: http://localhost:11434   # optional — falls back to cforch.ollama_url

  products:
    - id: peregrine
      name: Peregrine
      icon: "🦅"
      description: Job search assistant — live job listings
      base_url: http://localhost:8601
      health_path: /api/jobs/counts
      sample_endpoint: /api/jobs?status=pending&limit=5
      text_fields: [title, company, description]
      prompt_template: "Analyze this job listing and identify the key requirements, must-have skills, and any culture signals that would help tailor an application:\n\n{text}"

    - id: osprey
      name: Osprey
      icon: "📞"
      description: Gov't hold-line automation — recent call records
      base_url: http://localhost:8520
      health_path: /api/health
      sample_endpoint: /api/calls/recent
      text_fields: [agency, issue, notes]
      prompt_template: "Draft a clear, professional follow-up letter for this government hold-line call. Include what was discussed, what action the agency committed to, and a polite deadline for response:\n\n{text}"

    - id: linnet
      name: Linnet
      icon: "🐦"
      description: Real-time tone annotation — Elcor-style subtext for ND users
      base_url: http://localhost:8522
      health_path: /health
      sample_endpoint: /samples
      text_fields: [text, context]
      prompt_template: "Annotate the emotional tone and subtext of the following text using explicit Elcor-style markers (e.g. [SINCERELY], [UNCERTAIN], [FRUSTRATED]). Identify implied emotions, potential sarcasm, and any ambiguity that might be misread by neurodivergent readers:\n\n{text}"

    - id: kiwi
      name: Kiwi
      icon: "🥝"
      description: Pantry tracker
      base_url: http://localhost:8511
      sample_endpoint: /api/inventory
      text_fields: [name, category, notes]
      prompt_template: "Describe this pantry item and estimate how best to use it:\n\n{text}"

    - id: snipe
      name: Snipe
      icon: "🎯"
      description: eBay trust scoring
      base_url: http://localhost:8509
      sample_endpoint: /api/listings
      text_fields: [title, description, seller_info]
      prompt_template: "Evaluate the trustworthiness of this listing and flag any red flags:\n\n{text}"

    - id: pagepiper
      name: Pagepiper
      icon: "📄"
      description: "PDF/rulebook RAG tool: page-level text chunks"
      base_url: http://localhost:8511
      health_path: /api/health
      sample_endpoint: /api/library
      chunk_endpoint: /api/library/sample-chunks?limit=50   # requires pagepiper#6
      text_fields: [title]
      prompt_template: "Summarize the key rules described in this passage:\n\n{text}"

# ── Log corpus (Turnstone training data) ──────────────────────────────────────
corpus:
  # Directory containing pipeline JSONL log files to ingest (pull-side).
  # Files named <script>_<ts>.jsonl; one structured record per line.
  # POST /api/corpus/pipeline-ingest walks this dir and imports new files.
  # NFS-mounted on both Heimdall and Sif at /Library/Assets/
  pipeline_ingest_dir: /Library/Assets/logs/pipeline/

  # Turnstone push sources (consent-gated, token-authenticated).
  # sources:
  #   - token: "your-bearer-token"
  #     source_host: "node.local"
  #     owner: YourName
  #     consent_date: "2026-05-17"
  #     consent_method: signal_chat

# ── Embedding model comparison harness ────────────────────────────────────────
embed_bench:
  # ollama_url: http://localhost:11434   # optional; falls back to cforch.ollama_url
  # top_k: 5                            # default hits per model per query