the URL must start with the protocol postgresql

Not specified by the original author.

TL;DR: Your DATABASEURL does not start with postgresql:// or postgres://. Copy the full connection string from your database provider, including the scheme prefix, and set it as DATABASEURL. The value is malformed, not missing.

error: Error validating datasource `db`: the URL must start with the protocol `postgresql://` or `postgres://`.

Steps

1. Print the value Prisma sees (mask the password when sharing):

   node -e "const u=new URL(process.env.DATABASE_URL); console.log(u.protocol+'//'+u.host+u.pathname)"

   Expected: postgresql://YOUR_HOST:5432/mydb. If node throws 'Invalid URL', the value is malformed.

2. Fix the common shapes:

   • Missing scheme: YOUR_HOST:5432/mydb becomes postgresql://YOUR_HOST:5432/mydb
   • Wrong provider: a mysql:// URL against a provider = "postgresql" schema fails; match the scheme to the provider
   • SQLite path in a Postgres schema: file:./dev.db fails; point at a real Postgres or change the provider

   Expected: the value now parses as a URL with the right protocol.

3. Set it where the failing command runs (CI secret, container env, or local shell) and re-run:

   npx prisma validate

   Expected: The schema is valid instead of the P1012 protocol error.

When to use

• The exact message is the P1012 protocol error on validate, generate, migrate, or db push
• The provider in the schema is postgresql but the URL looks like anything else

When not to use

• Environment variable not found: DATABASE_URL: the variable is missing entirely, a different fix
• P1000/P1001: the URL parses fine but the server rejects you or is unreachable

Compatibility

Prisma 2.x through 7.x. The message text is the same across versions.

Variant phrasing: the URL must start with the protocol file:``

Same shape, SQLite provider. Your URL must start with file: (for example file:./dev.db) when the provider is sqlite.

Variant phrasing: protocol error after switching providers

Changing provider in the schema without changing the URL is the classic trigger. Update both together.

Why it happens

Prisma validates the datasource URL against the provider before connecting. A URL with no scheme, the wrong scheme, or a leftover from another provider fails this check immediately, which is why the fix is always in the value, never in Prisma itself.

Edge cases

• Quoting: some shells strip characters from unquoted URLs; quote the assignment.
• Some providers issue postgres:// and some postgresql://; Prisma accepts both for the postgresql provider.
• On Prisma 7 the same check runs against prisma.config.ts datasource url; fix the value there too.

A Railway deployment failed with P1012: 'Error validating datasource db: the URL must start with the protocol postgresql:// or postgres://.' The project's fix guide traces it to a DATABASEURL that was empty or missing the scheme prefix. The fix that worked: open the Postgres service credentials, copy the exact connection string including the postgresql:// prefix, and set it as DATABASEURL on the app service. After that, validation passed and the deploy went green. Source: https://github.com/ronaldo9991/collabotree/blob/HEAD/RAILWAYDATABASEURL_FIX.md

Not specified by the original author.

TL;DR

NOTE FROM MAINTAINERS The way to use @vercel/postgres locally is documented here: https://vercel.com/docs/storage/vercel-postgres/local-development --- I thought I could use Vercel Postgres with a local DB for offline development with a custom connection string that points to a local DB......

Steps

1. @vercel/postgres connects through websockets (it is a Neon client), not plain TCP, so pointing it at a local Postgres fails. Two ways around it: follow the documented local-development flow for Vercel Postgres, or run Neon's wsproxy between the client and your local Postgres on port 5432, which one user confirmed works. Also double-check POSTGRES_URL for typos: one reporter's whole issue was a malformed env value.

When to use

You are seeing this: @vercel/postgres connects through websockets (it is a Neon client), not plain TCP, so pointing it at a local Postgres fails. Use this skill when you run into "Trying to use a local DB with Vercel Postgres fails due to strict pool URL check".

When not to use

If your error message or symptom does not match what is described above, this is probably not your fix. Search for your exact error text instead of forcing this one to fit.

Versions

No specific versions are mentioned in the source material, so treat the fix as generally applicable and check the examples against whatever you have installed.

Why this happens

The original report does not dig into a root cause. It documents the symptom and the fix that resolved it.

Edge cases

• NOTE FROM MAINTAINERS The way to use @vercel/postgres locally is documented here: https://vercel.com/docs/storage/vercel-postgres/local-development --- I thought I could use Vercel Postgres with a local DB for offline development with a custom connection string that points to a local DB......
• Also double-check POSTGRES_URL for typos: one reporter's whole issue was a malformed env value.

Sources

• https://vercel.com/docs/storage/vercel-postgres/local-development

In gh:vercel/storage#123 (https://github.com/vercel/storage/issues/123), the reporter ran into this: NOTE FROM MAINTAINERS The way to use @vercel/postgres locally is documented here: https://vercel.com/docs/storage/vercel-postgres/local-development --- I thought I could use Vercel Postgres with a local DB for offline development with a custom connection string that points to a local DB... ... The verified resolution: @vercel/postgres connects through websockets (it is a Neon client), not plain TCP, so pointing it at a local Postgres fails. Two ways around it: follow the documented local-development flow for Vercel Postgres, or run Neon's wsproxy between the client and your local Postgres on port 5432, which one user confirmed works. Also double-check POSTGRES_URL for typos: one reporter's whole issue was a malformed env value.

Not specified by the original author.

TL;DR

Contributor Vidit-Ostwal identified a bug in crewai 0.121.0 specific to.ipynb notebooks: reporter fzozyurt confirmed 'the problem was solved when I tried it in Python' (.py files instead of the notebook).

Steps

1. Issue crewAIInc/crewAI#2746 (closed, 21 comments): ### Description Hello, I call Knowledge with JSONKnowledgeSource(filepaths=["log.json"]) in CrewAI, and I use Huggingface as Embedding, but I get an error like the one below. When I use StringKnowledgeSource instead of JSON, it does not give this error and does the operation. I think there is a problem with JSON. The error I get is: [2025-04-26 14:27:54][ERROR]: Failed to upsert documents: Request URL is missing an 'http://' or 'https://' protocol. in upsert. Code: ` jsonsource = JSONKnowledgeSource(filepaths=["log.json"]) crew = Crew( agents=[agent], tasks=[sampletask], process=Process.se
2. Contributor Vidit-Ostwal identified a bug in crewai 0.121.0 specific to .ipynb notebooks: reporter fzozyurt confirmed 'the problem was solved when I tried it in Python' (.py files instead of the notebook). If hitting the upsert error, run as a plain Python script or upgrade past 0.121.0; also verify the embedder config uses an api_url with http(s):// for HuggingFace TEI embeddings.

When to use

You are seeing this: I think there is a problem with JSON. Use this skill when you run into "[BUG] JSONKnowledgeSource Error Failed to upsert documents: Request URL is missing an 'http://' or 'https://' protocol".

When not to use

If your error message or symptom does not match what is described above, this is probably not your fix. Search for your exact error text instead of forcing this one to fit.

Versions

Versions mentioned in the source: crewai 0.121.0. If you are on something much newer or older, the details may have shifted.

Why this happens

The original report does not dig into a root cause. It documents the symptom and the fix that resolved it.

Sources

• http://'
• https://'

In gh:crewAIInc/crewAI#2746 (https://github.com/crewAIInc/crewAI/issues/2746), someone reported: Issue crewAIInc/crewAI#2746 (closed, 21 comments): ### Description Hello, I call Knowledge with JSONKnowledgeSource(filepaths=["log.json"]) in CrewAI, and I use Huggingface as Embedding, but I get an error like the one below. When I use StringKnowledgeSource instead of JSON, it does not give this error and does the operation. I think there is a problem with JSON. The error I get is: [2025-04-26 14:27:54][ERROR]: Failed to upsert documents: Request URL is missing an 'http://' or 'https://' protocol. in upsert. Code: ` jsonsource = JSONKnowledgeSource(filepaths=["log.json"]) crew = Crew( agents=[agent], tasks=[sampletask], process=Process.se The verified resolution from the thread: Contributor Vidit-Ostwal identified a bug in crewai 0.121.0 specific to .ipynb notebooks: reporter fzozyurt confirmed 'the problem was solved when I tried it in Python' (.py files instead of the notebook). If hitting the upsert error, run as a plain Python script or upgrade past 0.121.0; also verify the embedder config uses an api_url with http(s):// for HuggingFace TEI embeddings.