AWS Bedrock Mistral API

Same idea as the Bedrock Claude node, pointed at Mistral instead. It's the connection node that lets you call Mistral's models from ComfyUI through Amazon Bedrock, using your AWS credentials - no separate Mistral account needed. It outputs an LLM_API handle for the Chat and Completion nodes. If your infrastructure already lives in AWS and you want an open-weight-lineage model rather than Claude or GPT, this is the door.

Mistral's a solid, cost-effective choice for text tasks - prompt rewriting, caption expansion, structured extraction - the kind of LLM work you'd bolt onto an image pipeline. Through Bedrock it bills and governs the same way as the rest of your AWS usage.

How it fits together

The familiar three-node LLM pattern: this API node for AWS credentials, an LLM API Config for model and settings, and an LLM Chat or LLM Completion to actually run the call. This node only does the AWS authentication.

The inputs and outputs

• aws_access_key_id / aws_secret_access_key (STRING) - your AWS credentials.
• aws_session_token (STRING) - for temporary STS credentials (assumed roles, SSO). Leave it if you're on long-lived keys.
• region - which AWS region to call (default us-east-1, nine total). Bedrock model access is region-scoped, so this has to line up with where you enabled Mistral.
• llm_api (output) - the connection handle for the Chat/Completion nodes.

Note there's no version field here, unlike the Claude node - one fewer thing to set.

Installing it

ComfyUI Manager → search comfyui-art-venture → install → restart. Or:

cd ComfyUI/custom_nodes
git clone https://github.com/sipherxyz/comfyui-art-venture

then restart ComfyUI. Pre-installed on comfy.icu. Uses the boto3 AWS SDK from the pack's requirements.

Common issues

Missing-node red means art-venture isn't installed - Manager → "Install Missing Custom Nodes."

The Bedrock gotchas are identical to the Claude node's, so learn them once. You must request access to the specific Mistral model in the Bedrock console, per region, before it responds - nothing is enabled by default. The region has to match where you turned the model on. Your IAM policy needs bedrock:InvokeModel. If the Chat node returns access-denied or "model not found," it's almost always one of those three, not the graph. Confirm the model is enabled in your selected region first.

And the standing credential warning: AWS keys typed into this node get saved into the workflow file. Never share or embed a workflow that still has live AWS secrets in it - scrub them, and prefer short-lived STS session tokens to long-lived keys. A leaked cloud credential is a real incident, not an inconvenience.