wrangler.toml.example

name = "vectorize-mcp-worker-python"
main = "src/entry.py"
compatibility_date = "2026-02-12"
compatibility_flags = ["python_workers"]

[ai]
binding = "AI"

# RAG Index (384 dimensions - BGE text embeddings)
[[vectorize]]
binding = "VECTORIZE"
index_name = "mcp-knowledge-base"

[[d1_databases]]
binding = "DB"
database_name = "mcp-knowledge-db"
database_id = "YOUR_DATABASE_ID_HERE"  # Replace with your actual D1 database ID

# Service binding to the multimodal worker (internal-only)
[[services]]
binding = "MULTIMODAL"
service = "multimodal-pro-worker"

[observability]
enabled = true

# Uncomment to enable verbose debug logging (visible via wrangler tail --format=json)
# [vars]
# DEBUG_LOGGING = "true"

# Set your secrets using: pywrangler secret put <NAME>
# Never commit actual secret values to git
#   API_KEY          - Bearer token for authenticated endpoints
#   INTERNAL_SECRET  - Shared secret between main worker and multimodal worker
#   DEBUG_LOGGING    - Set to "true" to enable debug logs (can also use [vars] above)