🚨 Production Impact: Repeated 429 Resource exhausted Errors Despite Being Within Quota (Paid Tier 1)

[wo-yashsinghvi]

Hi Team,

We are facing persistent 429 Resource exhausted errors while calling the Gemini model via Vertex AI, and this is impacting a production application.

❗ Error Observed

429 Resource exhausted. Please try again later.
Please refer to https://cloud.google.com/vertex-ai/generative-ai/docs/error-code-429

🔍 Context

• Model: Gemini Flash 2.0
• Environment: Production
• Billing: Paid Tier 1 (active billing account)
• Usage: Well within the configured quota limits
• Integration: Using langchain-google package with Gemini (Vertex AI)

To be very clear:
➡️ This does not appear to be a LangChain issue.
➡️ The error is coming directly from Vertex AI / Gemini.

🚨 Business Impact

• This is a live production system
• End customers are being affected
• Requests are failing intermittently without any quota breach
• There is no clear signal why the resource is considered “exhausted”

❓ What We Need Help With

• Why are 429 Resource exhausted errors occurring even when usage is within quota?
• Are there any hidden limits, regional constraints, or backend throttling mechanisms for Gemini Flash 2.0?
• Is there an ongoing issue or degradation with the service?

If you need additional logs, project details, request IDs, or configuration information, please let us know—we’re happy to provide it immediately.

This issue is blocking business operations, so an urgent review and fix would be greatly appreciated.

[wo-yashsinghvi]

@janasangeetha Could you please help me with this issue?

[janasangeetha]

Hey @wo-yashsinghvi
Thanks for reaching out!
I have added you to a private Gist where you can share your project details. Kindly share the same, so we can check this issue.

[wo-yashsinghvi]

@janasangeetha Done

[miguelmanlyx]

@wo-yashsinghvi
Hi that looks like a Retries / Rate limits issue. Have you tried running npx ai-patch doctor to sanity-check 429s, backoff/retry behavior, and retry storm patterns?

[wo-yashsinghvi]

@miguelmanlyx I’ve already verified this, and I’m still receiving a 429 error despite being within the documented limits. Since this is a paid service, it’s unclear why the request is being rate-limited.

@janasangeetha I’ve already provided the project details through the private Gist. Could you please share an update on the status?

[jkeuleya]

I can see that we're a lot of people complaining about that since several days. I don't use this repo but I have 429 on almost 95% of my API requests. Few weeks ago, it was working very well.
Comes from Google and nobody gives any answer.