[Feature]: quick-start guide for Bedrock Anthropic

[codekiln]

Description

As a user in an enterprise that uses AWS, I would like the ability to use AWS role authentication initiate Bedrock Anthropic API calls using opencommit.

[commit] $ opencommit --provider anthropic 
Applying migration 00_use_single_api_key_and_url
│
└  Failed to apply migration 00_use_single_api_key_and_url: Error: Migration failed, set AI provider first. Run "oco config set OCO_AI_PROVIDER=<provider>", where <provider> is one of: ollama, openai, anthropic, gemini, azure, test, flowise, groq, mistral, mlx, deepseek, openrouter

Suggested Solution

1. include instructions in README on which environment variables to set to utilize Anthropic models through AWS Bedrock
2. in onboarding validation, provide bedrock-anthropic as a provider

Alternatives

Using an alternate provider - some enterprises won't allow other providers.

Additional Context

No response

[di-sukharev]

this needs to be investigated, opencommit works as a github action, please look how it is set up, would this somehow help you put env variables similarly in you environment?

[codekiln]

@di-sukharev I don't think it's quite as simple as just adding some env variables. Supporting it would likely entail opencommit using https://github.com/aws/aws-sdk-js-v3 to support default bedrock calls, at which point it could be configured with environment variables (such as AWS_PROFILE which point to ~/.aws/config). I haven't tested this example (from o4-mini).

import { BedrockRuntimeClient, InvokeModelCommand } from "@aws-sdk/client-bedrock-runtime";
import { Readable } from "stream";

// Helper to convert a Readable stream to string
async function streamToString(stream: Readable): Promise<string> {
  const chunks: Uint8Array[] = [];
  for await (const chunk of stream) {
    chunks.push(typeof chunk === "string" ? Buffer.from(chunk) : chunk);
  }
  return Buffer.concat(chunks).toString("utf-8");
}

async function generateCommitMessage(prompt: string): Promise<string> {
  // Make sure these env vars are set:
  // AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, AWS_SESSION_TOKEN (if using temporary creds), AWS_REGION
  const client = new BedrockRuntimeClient({
    region: process.env.AWS_REGION,
  });

  const command = new InvokeModelCommand({
    modelId: "anthropic.claude-v1",       // or your target Anthropic model alias
    contentType: "application/json",
    accept: "application/json",
    inputStream: Buffer.from(
      JSON.stringify({
        prompt,                           // e.g. "Write a concise git commit message for:\n…"
        max_tokens_to_sample: 200,        // adjust as needed
        temperature: 0.2
      })
    ),
  });

  const response = await client.send(command);
  if (!response.body) {
    throw new Error("Empty response from Bedrock");
  }

  const text = await streamToString(response.body as Readable);
  // Bedrock returns JSON with a `completion` field:
  const { completion } = JSON.parse(text);
  return completion.trim();
}

// Example usage:
(async () => {
  const diffSummary = `Modified typescript/opencommit.ts:
  - added support for AWS Bedrock Anthropics
  - refactored env var handling
  `;
  const commitMessage = await generateCommitMessage(
    `Generate a concise git commit message for the following diff:\n${diffSummary}`
  );
  console.log("AI-generated commit message:", commitMessage);
})();

For teams that have the bandwidth to set up an OpenAI-compatible proxy endpoint for bedrock in their IaC (see also https://github.com/aws-samples/bedrock-access-gateway), it might make it easier for opencommit to support them, but that's likely a small subset of bedrock users IMO.