I keep getting 404 errors when using a Google API key | Friends of the Crustacean 🦞🤝 | Page 1

kindred hawk Feb 27, 2026, 4:44 AM

#

LLM error: { "error": { "code": 400, "message": "Invalid JSON payload received. Unknown name "functionResponse" at 'contents[8].parts[0]': Cannot find field.\nInvalid JSON payload received. Unknown name "functionResponse" at '...

maiden lakeBOT Feb 27, 2026, 4:44 AM

#

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> ✅ Mark Solution

magic sequoiaBOT Feb 27, 2026, 4:48 AM

#

That Unknown name "functionResponse" at 'contents[...].parts[...]' error is coming from Google rejecting tool-call results in the request body.

In OpenClaw, when the model calls a tool, the next turn sends the tool result back as a Gemini functionResponse part. You’ll see this error when the endpoint you’re hitting doesn’t support that field — most commonly because you’re pointing at the wrong Gemini REST version (or a proxy) that only supports plain text parts.

What to check / fix

Check if you overrode the Google base URL

If you have models.providers.google.baseUrl set to https://generativelanguage.googleapis.com/v1 → switch it to v1beta:
https://generativelanguage.googleapis.com/v1beta
Or simplest: remove your entire models.providers.google override and just use the built-in google provider with GEMINI_API_KEY.

Make sure you’re using the Gemini API-key provider (google), not Vertex

google = Gemini API key (GEMINI_API_KEY)
google-vertex = Vertex AI (ADC / service account), different setup

Restart the gateway after changes

#

openclaw gateway restart

Quick debug info that’ll confirm it fast

Paste these (redact keys):

openclaw --version
openclaw models status
openclaw config get agents.defaults.model
openclaw config get models.providers.google

The models.providers.google block is the big one—if it shows baseUrl: .../v1 (or some proxy URL), that’s almost certainly the cause.

kindred hawk Feb 27, 2026, 4:56 AM

#

its not working, I changed the url to v1beta on the config file, should I just completely remove that section so it isn't in config json

magic sequoiaBOT Feb 27, 2026, 4:57 AM

#

Yeah — unless you intentionally run Gemini through a custom proxy/base URL, you should delete the whole models.providers.google override (or at least delete the google: { ... } block).

OpenClaw already has a built-in google provider config for Gemini; you typically only need to supply the key (GEMINI_API_KEY). Your current error (Unknown name "functionResponse") is a strong hint you’re still hitting an endpoint/schema that isn’t the real Gemini Generative AI API (or the gateway is still using the old config).

Do this exact sequence:

Remove models.providers.google from your config (so it falls back to the built-in provider).
Put your key in one of these places:

GEMINI_API_KEY in ~/.openclaw/.env (common when running as a daemon/service), or
env: { GEMINI_API_KEY: "..." } in your openclaw.json

Restart:

openclaw gateway restart

Verify:

#

openclaw models status

If it still errors, paste:

openclaw --version
openclaw gateway status
openclaw models status

openclaw gateway status is important because it often reveals you edited the wrong config file (CLI vs service config), so your change never actually took effect.

Docs for reference:

kindred hawk Feb 27, 2026, 5:09 AM

#

it is still erroring

#

here is the log

#

Service: LaunchAgent (loaded)
File logs: /tmp/openclaw/openclaw-2026-02-27.log
Command: /opt/local/bin/node /opt/local/lib/node_modules/openclaw/dist/index.js gateway --port 18789
Service file: ~/Library/LaunchAgents/ai.openclaw.gateway.plist
Service env: OPENCLAW_GATEWAY_PORT=18789

Config (cli): ~/.openclaw/openclaw.json
Config (service): ~/.openclaw/openclaw.json

Gateway: bind=loopback (127.0.0.1), port=18789 (service args)
Probe target: ws://127.0.0.1:18789
Dashboard: http://127.0.0.1:18789/
Probe note: Loopback-only gateway; only local clients can connect.

Runtime: running (pid 30137)
RPC probe: ok

Listening: 127.0.0.1:18789
Troubles: run openclaw status
Troubleshooting: https://docs.openclaw.ai/troubleshooting

magic sequoiaBOT Feb 27, 2026, 5:11 AM

#

If it’s still throwing Unknown name "functionResponse" after you changed/removed the Google provider block, that basically means one of these is true:

the Gateway is not using the config file you edited (super common: daemon/service uses a different config), or
it’s still hitting a Gemini endpoint variant/proxy that doesn’t accept tool-result parts (e.g. the /openai compatibility route, or some proxy URL).

Can you paste the output of these (redact keys)? They’ll tell us which one it is immediately:

openclaw --version
openclaw gateway status --deep
openclaw config get models.providers.google
openclaw models status --json

What I’m looking for in those outputs:

In gateway status --deep: whether Config (cli) and Config (service) differ. If they differ, you edited the wrong one.
In models status --json: the actual baseUrl the running gateway is using for google. It must be the native Gemini API base URL (ends in /v1beta) and must NOT include /openai.

#

If you want a quick “sanity reset” while you grab outputs:

ensure models.providers.google is gone
put GEMINI_API_KEY in ~/.openclaw/.env (if you run as a service/daemon)
then:

openclaw gateway restart

then try again (ideally in a fresh chat session with /new, just to avoid replaying a broken tool-turn history)

Docs reference: https://docs.openclaw.ai/gateway/troubleshooting and https://docs.openclaw.ai/concepts/model-providers

#

The snippet you pasted is gateway status (service healthy, config path matches) — good news, but it doesn’t include the actual Gemini request URL / provider config that’s causing functionResponse to be rejected.

Next step: we need to confirm what the running gateway thinks the google provider baseUrl + api are.

Please paste these outputs (redact keys/tokens):

openclaw config get agents.defaults.model
openclaw models status --json

If --json is huge, just search within it and paste the google part (baseUrl + api). Easiest:

openclaw models status --json | grep -n \"\\\"google\\\"\" -n | head -60

#

Also paste the actual error lines from the file log it pointed you to:

grep -n "functionResponse" /tmp/openclaw/openclaw-2026-02-27.log | tail -n 20
grep -n "Invalid JSON payload received" /tmp/openclaw/openclaw-2026-02-27.log | tail -n 20

What we’re trying to catch:

Is the baseUrl really https://generativelanguage.googleapis.com/v1beta (no /openai, no Vertex aiplatform.googleapis.com, no proxy)?
Is the provider api actually google-generative-ai (and not something mismatched)?
Sometimes the config edit is correct, but the gateway is still using an old generated models.json / cached provider entry — models status --json tells the truth.

Once you paste those, I can tell you exactly what to change (and where).

kindred hawk Feb 27, 2026, 5:24 AM

#

d"}","1":"embedded run agent end: runId=172dfdb9-dfd3-421a-b97f-ca2d87ad16f9 isError=true error=LLM error: {\n "error": {\n "code": 400,\n "message": "Invalid JSON payload received. Unknown name \"functionResponse\" at 'contents[8].parts[0]': Cannot find field.\nInvalid JSON payload received. Unknown name \"functionResponse\" at 'contents[10].parts[0]': Cannot find field.\nInvalid JSON payload received. Unknown name \"systemInstruction\": Cannot find field.\nInvalid JSON payload received. Unknown name \"tools\": Cannot find field.",\n "status": "INVALID_ARGUMENT",\n "details": [\n {\n "@type": "type.googleapis.com/google.rpc.BadRequest",\n "fieldViolations": [\n {\n "field": "contents[8].parts[0]",\n "description": "Invalid JSON payload received. Unknown name \"functionResponse\" at 'contents[8].parts[0]': Cannot find field."\n },\n {\n "field": "contents[10].parts[0]",\n "description": "Invalid JSON payload received. Unknown name \"functionResponse\" at 'contents[10].parts[0]': Cannot find field."\n },\n {\n "description": "Invalid JSON payload received. Unknown name \"systemInstruction\": Cannot find field."\n },\n {\n "description": "Invalid JSON payload received. Unknown name \"tools\": Cannot find field."\n }\n ]\n }\n ]\n }\n}","_meta":{"runtime":"node","runtimeVersion":"22.22.0","hostname":"unknown","name":"{"subsystem":"agent/embedded"}","parentNames":["openclaw"],"date":"2026-02-27T05:10:39.290Z","logLevelId":4,"logLevelName":"WARN","path":{"fullFilePath":"file:///opt/local/lib/node_modules/openclaw/dist/subsystem-DypCPrmP.js:1170:14","fileName":"subsystem-

magic sequoiaBOT Feb 27, 2026, 5:27 AM

#

That extra bit in the error (Unknown name "systemInstruction" + Unknown name "tools") is the giveaway: you’re still hitting a Gemini endpoint/schema that doesn’t support tool calling (that’s why it also rejects functionResponse).

Even if you changed/removed models.providers.google in ~/.openclaw/openclaw.json, OpenClaw writes/merges provider overrides into a generated models registry file, and removing the config section does not delete the old entry.

Fix (most common)

Check if you have a stale Google provider in the generated registry:

ls -la ~/.openclaw/agents/main/agent/models.json
grep -n "generativelanguage\\|/v1\\b\\|v1beta\\|google" ~/.openclaw/agents/main/agent/models.json | head -60

If you see Google pointing at /v1 (not /v1beta), either edit it or (simpler) move it aside so it gets regenerated:

mv ~/.openclaw/agents/main/agent/models.json ~/.openclaw/agents/main/agent/models.json.bak
openclaw gateway restart

#I keep getting 404 errors when using a Google API key

What to check / fix

Quick debug info that’ll confirm it fast

Fix (most common)