Skip to content

docs(forest-runtime): document Amazon Bedrock as an AI provider - #33

Open
Scra3 wants to merge 3 commits into
mainfrom
docs/forest-runtime-bedrock
Open

docs(forest-runtime): document Amazon Bedrock as an AI provider#33
Scra3 wants to merge 3 commits into
mainfrom
docs/forest-runtime-bedrock

Conversation

@Scra3

@Scra3 Scra3 commented Sep 11, 2026

Copy link
Copy Markdown
Member

Companion to ForestAdmin/agent-nodejs#1897, which adds AI_PROVIDER=bedrock to the self-hosted executor.

Bedrock is the one provider that takes no AI_API_KEY — credentials come from the AWS chain, and a key set alongside it fails at startup rather than being ignored. The previous wording made the three variables sound inseparable, so the note and the table both needed adjusting, not just a new row.

Adds a short Amazon Bedrock subsection with the required AWS_REGION, the bedrock:InvokeModel permission, and a copy-pasteable example.

Merge after the agent-nodejs PR ships.

🤖 Generated with Claude Code

Note

Document Amazon Bedrock as an AI provider in forest-runtime

Updates the AI provider configuration docs to include bedrock, adjusts the API-key wording so it is not required for Bedrock, and revises the all-or-nothing configuration note. Adds a Bedrock section covering supported Anthropic model families, model and inference-profile identifiers, regional residency guidance, AWS credential-chain behavior, required IAM permission, role-assumption setup options, the AWS_REGION setting, and a Bedrock environment-variable example.

Macroscope summarized 73f4231.

AI_PROVIDER now accepts bedrock, for customers who need their LLM calls to stay
inside their AWS account. It is the one provider that takes no AI_API_KEY:
credentials come from the AWS chain, and a key set alongside it fails at startup
rather than being ignored — worth stating, since the previous wording made the
three variables sound inseparable.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
@mintlify

mintlify Bot commented Sep 11, 2026

Copy link
Copy Markdown
Contributor

Preview deployment for your docs. Learn more about Mintlify Previews.

Project Status Preview Updated
forest 🟢 Ready View Preview Sep 14, 2026, 9:35 AM

💡 Tip: Enable Automations to automatically generate PRs for you.


```bash
AI_PROVIDER=bedrock
AI_MODEL=us.anthropic.claude-sonnet-4-6-v1:0

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🟠 High workflows/forest-runtime.mdx:237

The copy-paste configuration cannot invoke us.anthropic.claude-sonnet-4-6-v1:0 from AWS_REGION=eu-west-3: it is a US geo profile, and it is not the documented Sonnet 4.6 profile ID. Use the Paris-compatible eu.anthropic.claude-sonnet-4-6 profile, or pair the US profile with a supported US source Region.

Suggested change
AI_MODEL=us.anthropic.claude-sonnet-4-6-v1:0
AI_MODEL=eu.anthropic.claude-sonnet-4-6
🚀 Reply "fix it for me" or copy this AI Prompt for your agent:
In file @product/process/workflows/forest-runtime.mdx around line 237:

The copy-paste configuration cannot invoke `us.anthropic.claude-sonnet-4-6-v1:0` from `AWS_REGION=eu-west-3`: it is a US geo profile, and it is not the documented Sonnet 4.6 profile ID. Use the Paris-compatible `eu.anthropic.claude-sonnet-4-6` profile, or pair the US profile with a supported US source Region.

Comment thread product/process/workflows/forest-runtime.mdx
A customer's devops asked whether they can assume a role instead of providing
keys. They can — the SDK's own chain handles it — but the previous wording only
named task roles, profiles and static keys, so it read as if those were the
whole list.

Also warns that AWS_REGION is required even when the profile declares one: the
Bedrock client does not read the region from the profile, and that is the one
place where the chain does not behave as a reader would expect.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
| Static keys | `AWS_ACCESS_KEY_ID` and `AWS_SECRET_ACCESS_KEY`. |

<Warning>
`AWS_REGION` is required even when your profile already declares a `region` — the Bedrock client does not read the region from the profile. Forest Runtime refuses to start without it rather than failing on the first AI step.

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🟡 Medium workflows/forest-runtime.mdx:245

The documentation incorrectly says AWS_REGION is mandatory, even though the executor also accepts AWS_DEFAULT_REGION; users with only the standard AWS variable are therefore told their valid setup will be rejected. Document both accepted variables here.

Suggested change
`AWS_REGION` is required even when your profile already declares a `region` — the Bedrock client does not read the region from the profile. Forest Runtime refuses to start without it rather than failing on the first AI step.
`AWS_REGION` or `AWS_DEFAULT_REGION` is required even when your profile already declares a `region` — the Bedrock client does not read the region from the profile. Forest Runtime refuses to start without either variable rather than failing on the first AI step.
🚀 Reply "fix it for me" or copy this AI Prompt for your agent:
In file @product/process/workflows/forest-runtime.mdx around line 245:

The documentation incorrectly says `AWS_REGION` is mandatory, even though the executor also accepts `AWS_DEFAULT_REGION`; users with only the standard AWS variable are therefore told their valid setup will be rejected. Document both accepted variables here.

Bedrock resells hundreds of models and the runtime only accepts Claude Sonnet,
Haiku and Opus, so the page needs to say which ids work rather than leave the
reader to try one and get a startup refusal.

Also warns against the global. prefix: it routes worldwide, which defeats the
data residency that sends most customers to Bedrock in the first place.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant