Hi,

Sharing two exact errors I hit today building on Archestra: with *Cerebras, a model call failed with Status: 402, Type: unknown_api_error, Message: "402 status code (no body)" — no further detail in the response. With Groq* (Llama 3.3 70B), even in a fresh chat with a short prompt, I got Status: 413, Type: api_payload_too_large_error, Message: "Request body too large: 0.0 MB (limit 70 MB). Use a smaller attachment, or raise ARCHESTRA_API_BODY_LIMIT." — the reported size (0.0 MB) doesn't match the "too large" error. Posting in case others have hit the same and can point to a cause — happy to share full raw JSON for either if useful.

hey .. actually I got another issue .. after stopping the record and saving the prompt and so on when I try to download it it just says it will be exported once it's ready and you download it but nothing got out to be saved on my device and when I try to submit the thing it says sign in to github and once I press it another screen shows that I didn't define the prompt or whatever even tho that I chose the chat messages about the prompt and cut the recording to 30 secs ... is that normal and is it happening with all or I'm missing smth ?

Hi , I am trying to create an app, but the agent says it has scaffolded the app but it doesnt persist the files. Is anyone else facing this issue ?

When i am adding the gmail in mcp registry then i am getting this error any resolution to this ?

Heyy

Hitting persistent rate-limit issues while testing app builds in

Archestra — bounced between OpenRouter (empty responses, then a 402

credits error even though I only asked for a simple test app) and

Gemini free tier (RESOURCE_EXHAUSTED after just ~5 requests). Wondering

if there's a recommended provider/model combo for hackathon use that

has more reasonable free-tier limits for iterative app building, since

retrying a few times during testing burns through quota fast. Any

suggestions?

Hello, is there already a ready-made solution for using ArchestrA in a multi-tenant format? Specifically, we would like a single installation to serve multiple companies within the group. Is this possible yet?

Running into repeated LLM provider issues while building an app in

Archestra — tried OpenRouter (empty responses, then a 402 error saying

it requested 32,768 max_tokens but my free account can only afford

1,600-4,000 depending on the moment), Gemini free tier (rate-limited

after ~5 requests), and OpenAI (empty response with no clear reason).

The OpenRouter 32k max_tokens default seems like it might be worth

lowering for free-tier accounts, or exposing as an easy setting — right

now it silently fails instead of scaling down. Is there a way to

configure max_tokens per-model, or a recommended provider/model combo

that works reliably on free tiers for iterative app building?

Hi

Is archestra.app down right now? Getting repeated Cloudflare 522

"connection timed out" errors when trying to open apps I've built via

the chat — happening consistently across multiple different apps and

LLM providers over the last few hours, so it doesn't seem related to

my setup. App builds succeed in chat, but the preview link always

522s. Anyone else seeing this, or any workaround to view apps while

it's down? so any info helps.

Does the platform support multiple languages?

@user @user Again I faces the same issue?

Hi, been having this repeated challenge for hours now, even after changing models, see debug error (tried to de-auto-select model but not activating)

Your conversation is too long. Please start a new chat or remove some messages.

Agent: My Assistant

Model: 3cbf7cf2-a4d6-4f99-b73a-348801f7****

Provider: groq

Model source: auto-selected (best available)

Code: contexttoolong

Provider: groq

Status: 413

Type: apipayloadtoolargeerror

Message: Request too large for model llama-3.3-70b-versatile in organization org_01kxd8hjn4f7vtzgf55x2e48x9 service tier on_demand on tokens per minute (TPM): Limit 12000, Requested 35421, please reduce your message size and try again. Need more tokens? Upgrade to Dev Tier today at console.groq.com/settings/billing

Raw Error:

{

"url": "localhost/v1/groq/…/completions",

"statusCode": 413,

"isRetryable": false,

"responseBody": {

"error": {

"message": "Request too large for model llama-3.3-70b-versatile in organization org_01kxd8hjn4f7vtzgf55x2e48x9 service tier on_demand on tokens per minute (TPM): Limit 12000, Requested 35421, please reduce your message size and try again. Need more tokens? Upgrade to Dev Tier today at console.groq.com/settings/billing",

"type": "apipayloadtoolargeerror"

}

}

}

Session: c540c217-e47a-423c-a3a3-11998a0b718a

Trace: e8744659ee4413956dbe0ed71e30c13b

Span: 08758220ccd7eafe

new chat did not fix it

Hi, I only have a RuPay card, and both OpenRouter and OpenAI reject it with "This card type isn't accepted." Gemini now requires prepaid credits in my region. Is there a hackathon API key, sponsored credits, or another recommended provider that participants with RuPay cards can use?

Hi! I have a quick question regarding the submission. Are we allowed to include a voice-over narration in our demo video to explain the project and workflow, or is the submission expected to be screen recording only?

Thank you!

Hey guys! I’ve been testing out Archestra for a week. The main things I want to get working is to provide shared/curated skills with the rest of my team and provide them a way to share internal vibe-coded apps (I understand the feature is still in beta, I have some feedback that I’d love to share!). While testing some skills, in particular skills involving creating some HTML assets, I noticed that my agent uses archestra__run_command to write files. Problem with that is that I always hit a Error: command is too large (> 16384 bytes) error, and it forces the agent to unreliably find workarounds (once it tried rewriting the HTML file in small splits, forcing the content to be rewritten in small pages and links between them, another time it just zipped everything). I got the same when working with Apps, even more brutally given that, well, it’s all about apps. Is my agent misconfigured or is this a current limitation of Archestra?

hey .. actually I got another issue .. after stopping the record and saving the prompt and so on when I try to download it it just says it will be exported once it's ready and you download it but nothing got out to be saved on my device and when I try to submit the thing it says sign in to github and once I press it another screen shows that I didn't define the prompt or whatever even tho that I chose the chat messages about the prompt and cut the recording to 30 secs ... is that normal and is it happening with all or I'm missing smth ?

Hi! I'm blocked by what appears to be an Archestra issue.

In a fresh session, even sending "Hello" fails with:

Groq reports ~40,000 requested tokens despite the prompt being only a few characters.

Example:

Has anyone seen this before or is there a way to clear/reset the app context?

When starting a conversation with a model like Gemini, for example, can't I switch to another one, like OpenAI's?