feat: add Amazon Bedrock as inference provider by naelmohammad · Pull Request #8832 · NousResearch/hermes-agent

naelmohammad · 2026-04-13T06:41:18Z

Add Amazon Bedrock as a first-class inference provider using the Converse API via boto3. Supports three authentication methods: Bedrock API keys (Bearer tokens), AWS profiles (~/.aws/credentials), and AWS access key pairs.

agent/bedrock_adapter.py — Converse API adapter (message conversion, tool schema translation, response normalization)
hermes_cli/auth.py — ProviderConfig, _resolve_bedrock_base_url(), region-aware endpoint construction, provider aliases
hermes_cli/providers.py — HermesOverlay with bedrock_converse transport, aliases, labels
hermes_cli/models.py — Curated model list with global/eu/us inference profile IDs and bare model IDs
hermes_cli/config.py — AWS_BEARER_TOKEN_BEDROCK and AWS_BEDROCK_REGION env vars
hermes_cli/runtime_provider.py — bedrock_converse in valid API modes
agent/models_dev.py — PROVIDER_TO_MODELS_DEV mapping
New api_mode 'bedrock_converse' with full lifecycle support: init, switch_model, _build_api_kwargs, _interruptible_api_call, response validation, finish_reason extraction, response normalization
Safe vars() call in error logging (handles dict responses from boto3)
Dedicated _model_flow_bedrock() with interactive setup:
- Auth picker: existing API key, new API key, AWS profile, access keys
- Region selector with common regions + custom input
- Model selection from curated list
Persists api_mode=bedrock_converse in config.yaml
Added boto3 to pyproject.toml

Bedrock models use cross-region inference profile prefixes:

global. — works from any region
eu. — EU cross-region routing
us. — US cross-region routing
bare — single-region direct access

What does this PR do?

Related Issue

Fixes #
Adds Amazon Bedrock support as an inference provider using boto3 client.

Type of Change

Feature enhancement adding additional provider.

🐛 Bug fix (non-breaking change that fixes an issue)
✨ New feature (non-breaking change that adds functionality)
🔒 Security fix
📝 Documentation update
✅ Tests (adding or improving test coverage)
♻️ Refactor (no behavior change)
🎯 New skill (bundled or hub)

Changes Made

agent/bedrock_adapter.py — Converse API adapter (message conversion, tool schema translation, response normalization)
hermes_cli/auth.py — ProviderConfig, _resolve_bedrock_base_url(), region-aware endpoint construction, provider aliases
hermes_cli/providers.py — HermesOverlay with bedrock_converse transport, aliases, labels
hermes_cli/models.py — Curated model list with global/eu/us inference profile IDs and bare model IDs
hermes_cli/config.py — AWS_BEARER_TOKEN_BEDROCK and AWS_BEDROCK_REGION env vars
hermes_cli/runtime_provider.py — bedrock_converse in valid API modes
agent/models_dev.py — PROVIDER_TO_MODELS_DEV mapping

How to Test

launch hermes
select one of the anthropic models from Amazon Bedrock:
/model
Current: global.anthropic.claude-opus-4-6-v1 on Amazon Bedrock
Amazon Bedrock [--provider bedrock] (current):
global.anthropic.claude-opus-4-6-v1, global.anthropic.claude-sonnet-4-6, global.anthropic.claude-haiku-4-5-20251001-v1:0, anthropic.claude-opus-4-6-v1, anthropic.claude-sonnet-4-6, anthropic.claude-haiku-4-5-20251001-v1:0 (+4 more)

Checklist

Code

I've read the Contributing Guide
My commit messages follow Conventional Commits (fix(scope):, feat(scope):, etc.)
I searched for existing PRs to make sure this isn't a duplicate
My PR contains only changes related to this fix/feature (no unrelated commits)
I've run pytest tests/ -q and all tests pass
I've added tests for my changes (required for bug fixes, strongly encouraged for features)
I've tested on my platform:

Documentation & Housekeeping

I've updated relevant documentation (README, docs/, docstrings) — or N/A
I've updated cli-config.yaml.example if I added/changed config keys — or N/A
I've updated CONTRIBUTING.md or AGENTS.md if I changed architecture or workflows — or N/A
I've considered cross-platform impact (Windows, macOS) per the compatibility guide — or N/A
I've updated tool descriptions/schemas if I changed tool behavior — or N/A

For New Skills

This skill is broadly useful to most users (if bundled) — see Contributing Guide
SKILL.md follows the standard format (frontmatter, trigger conditions, steps, pitfalls)
No external dependencies that aren't already available (prefer stdlib, curl, existing Hermes tools)
I've tested the skill end-to-end: hermes --toolsets skills -q "Use the X skill to do Y"

Screenshots / Logs

Add Amazon Bedrock as a first-class inference provider using the Converse API via boto3. Supports three authentication methods: Bedrock API keys (Bearer tokens), AWS profiles (~/.aws/credentials), and AWS access key pairs. - agent/bedrock_adapter.py — Converse API adapter (message conversion, tool schema translation, response normalization) - hermes_cli/auth.py — ProviderConfig, _resolve_bedrock_base_url(), region-aware endpoint construction, provider aliases - hermes_cli/providers.py — HermesOverlay with bedrock_converse transport, aliases, labels - hermes_cli/models.py — Curated model list with global/eu/us inference profile IDs and bare model IDs - hermes_cli/config.py — AWS_BEARER_TOKEN_BEDROCK and AWS_BEDROCK_REGION env vars - hermes_cli/runtime_provider.py — bedrock_converse in valid API modes - agent/models_dev.py — PROVIDER_TO_MODELS_DEV mapping - New api_mode 'bedrock_converse' with full lifecycle support: init, switch_model, _build_api_kwargs, _interruptible_api_call, response validation, finish_reason extraction, response normalization - Safe vars() call in error logging (handles dict responses from boto3) - Dedicated _model_flow_bedrock() with interactive setup: - Auth picker: existing API key, new API key, AWS profile, access keys - Region selector with common regions + custom input - Model selection from curated list - Persists api_mode=bedrock_converse in config.yaml - Added boto3 to pyproject.toml Bedrock models use cross-region inference profile prefixes: - global. — works from any region - eu. — EU cross-region routing - us. — US cross-region routing - bare — single-region direct access

renlon · 2026-04-14T23:50:37Z

Thanks for working on Bedrock support — this is a much-needed feature. I've been working on an alternative implementation and wanted to share some observations that might be useful.

Core concern: Converse API vs Anthropic SDK

This PR uses the boto3 Converse API (bedrock-runtime.converse()), which is AWS's generic multi-model API. While this works for basic chat, it loses several Claude-specific features that the Anthropic SDK's AnthropicBedrock class preserves:

Feature	Converse API (this PR)	AnthropicBedrock SDK
Streaming	❌ Synchronous only (blocks until full response)	✅ Full streaming support
Prompt caching	❌ Not supported	✅ Same cache_control as native Anthropic
Reasoning/thinking	❌ Not exposed by Converse	✅ Full thinking block support
Adaptive thinking	❌ No	✅ Budget controls (low/medium/high/max)
Fast mode	❌ No	✅ Anthropic beta supported

For an agent framework where Claude is the primary model, losing streaming alone is a significant UX regression — users see nothing until the entire response is generated.

Alternative approach

The Anthropic Python SDK already ships with anthropic.AnthropicBedrock, which handles SigV4 signing internally and exposes the exact same messages.create() / messages.stream() interface as anthropic.Anthropic. This means the existing anthropic_messages API mode works as-is — no new message conversion, tool schema translation, or response normalization needed.

For bearer token auth (Bedrock API keys), AnthropicBedrock(api_key=bearer_token, aws_region=region) works directly.

This reduces the implementation from ~1,078 lines / 13 files to ~190 lines / 8 files (no new adapter file needed), with no new api_mode and no boto3 as a core dependency (only botocore at runtime, pulled in by the anthropic SDK's bedrock module).

Specific items

boto3 as a core dependency (pyproject.toml) — adds ~15MB (boto3 + botocore + s3transfer) to every install, even for users who don't use Bedrock. If sticking with Converse, this should be an optional extra ([bedrock]).
No streaming — bedrock_client.converse() is synchronous. The converse_stream() variant exists but isn't used here, and even that returns a different event structure than what _run_agent_loop() expects.
No prompt caching — The Converse API doesn't support Anthropic's cache_control breakpoints. For multi-turn agent conversations, this means ~4x higher input token costs on cached prefixes.
Missing CLAUDE_CODE_USE_BEDROCK support — Claude Code users expect this env var to activate Bedrock. The PR uses AWS_BEDROCK_REGION instead of AWS_REGION and doesn't support ANTHROPIC_MODEL / ANTHROPIC_SMALL_FAST_MODEL / DISABLE_PROMPT_CACHING.

Suggestion

Consider switching the underlying transport from boto3.converse() to anthropic.AnthropicBedrock. The interactive setup wizard and curated model list from this PR are valuable and could be kept — the change is primarily in agent/bedrock_adapter.py (which could be removed entirely) and the client construction path.

Happy to discuss further or share code from the alternative implementation.

alt-glitch · 2026-04-28T01:13:38Z

Related to #12979 (open, triaged) which implements full Bedrock Converse API support with cross-region inference profiles. Also related to closed #6845 and #7920 which attempted the same feature.

teknium1 · 2026-05-24T11:05:15Z

Closing as superseded by #13814.

Triage notes (high confidence):
Bedrock is already a first-class inference provider on main: agent/bedrock_adapter.py, agent/transports/bedrock.py, plugins/model-providers/bedrock/, tests/agent/test_bedrock_*. Merged via #13814 'feat: add BedrockTransport + wire all Bedrock transport paths' (2026-04-22).

Thanks for the contribution — the underlying problem this PR addresses has been resolved by the linked PR on current main. If you believe this was closed in error, please comment and we'll reopen.

(Bulk-closed during a CLI PR triage sweep.)

This was referenced Apr 14, 2026

feat: native AWS Bedrock provider via Converse API #7920

Closed

feat: add AWS Bedrock provider via Anthropic SDK #9957

Closed

alt-glitch mentioned this pull request Apr 28, 2026

feat(bedrock): support custom AWS Bedrock endpoint with bearer token auth #17162

Open

17 tasks

teknium1 closed this May 24, 2026

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: add Amazon Bedrock as inference provider#8832

feat: add Amazon Bedrock as inference provider#8832
naelmohammad wants to merge 1 commit into
NousResearch:mainfrom
naelmohammad:bedrock-integration

naelmohammad commented Apr 13, 2026 •

edited

Loading

Uh oh!

renlon commented Apr 14, 2026

Uh oh!

alt-glitch commented Apr 28, 2026

Uh oh!

teknium1 commented May 24, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Conversation

naelmohammad commented Apr 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What does this PR do?

Related Issue

Type of Change

Changes Made

How to Test

Checklist

Code

Documentation & Housekeeping

For New Skills

Screenshots / Logs

Uh oh!

renlon commented Apr 14, 2026

Core concern: Converse API vs Anthropic SDK

Alternative approach

Specific items

Suggestion

Uh oh!

alt-glitch commented Apr 28, 2026

Uh oh!

teknium1 commented May 24, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

naelmohammad commented Apr 13, 2026 •

edited

Loading