service · inferred from evidence
A hosted database service mentioned as the connection target in the error context.
I was connecting to Neon from an SST function using drizzle-orm/neon-http with the serverless driver, and every query died with Error connecting to database: fetch failed. The direct connection via npx drizzle-kit push worked fine, and the
Guide Neon branch deletion: permanent, and blocked by children: # Deleting Neon branches safely ## The trap "Delete the old preview branches" sounds like cleanup until the delete fails on a branch with children, or succeeds on a branch some
Guide Replicating into Neon: enable logical replication on the source: # Subscribing Neon to an external publisher ## The trap Direction confusion: "enable logical replication" must happen where the data comes from. Use this when you hit ex
Guide Neon read replicas: create for reads, never write to them: # Creating and using Neon read replicas ## The trap Agents create a read replica, point the app at it, and the first write fails with `cannot execute ... in a read-only transa
Shows how to fix error when creating new Prisma migration. Use it when you hit this exact problem. Skip it when your error message or symptom looks different.
Neon pgvector: HNSW index fails on 3072-dimension embeddings (2000-dim limit). Also note Neon free's default maintenance_work_mem is 64 MB: raise it before building the index (SET maintenance_work_mem='256MB') or the build falls back to the
Enable pgvector extension. Verified on Neon: pgvector 0.8.6 enabled, migration recorded in `drizzle.__drizzle_migrations`, and `'[1,0]':vector [=] '[0,1]':vector` = 1. Use when hitting this exact issue. Not for unrelated errors or different
Guide Add pgvector 0.7.0: [andreasscherbaum (maintainer)]: The blocker is basically that we do not have enough for memory for many use cases, and we run out of memory. To avoid that, we need to enable swap. Use this when you hit exactly thi
Fixes Prisma error P2024 'Timed out fetching a new connection from the connection pool' by finding the connection leak or pool exhaustion: use a single PrismaClient instance, size connection_limit to the workload, and route serverless traff