open-swe/deploy/seahaven/put-config.sh
Adam Moussa a4ed19ba61
Some checks failed
CI / Lint (push) Waiting to run
CI / Format check (push) Waiting to run
CI / Unit tests (push) Waiting to run
CI / Playwright E2E (push) Waiting to run
Build & publish app artifacts / Publish + deploy (dev) (push) Has been cancelled
Build & publish app artifacts / Publish + deploy (prod) (push) Has been cancelled
Infra CD / Infra CI (pre-deploy) (push) Has been cancelled
Infra CD / Deploy open-swe-dev (push) Has been cancelled
Infra CD / Deploy open-swe-prod (push) Has been cancelled
feat: migrate model providers to Bedrock (Claude) + Fireworks (everything else) (#62)
* feat: switch model providers to AWS Bedrock (Claude) and Fireworks (non-Claude)

Migrate off direct provider APIs: AWS Bedrock for Anthropic/Claude via the
cross-region inference profile us.anthropic.claude-opus-4-8, Fireworks AI for
all non-Claude models. Drop OpenAI (gpt-5.5) and Google (gemini-3.5-flash)
entirely. DEFAULT_MODEL_ID is now Bedrock Claude; all Fireworks models stay
freely selectable for the agent and reviewer graphs and via team/profile
defaults.

- pyproject: add langchain-aws (ChatBedrockConverse + boto3)
- options.py: Bedrock Claude entry + default; remove openai/google entries
- model.py: bedrock_converse provider_model_kwargs (effort -> thinking budget),
  region pin in make_model, bedrock<->fireworks fallback pairing, AWS_REGION/
  FIREWORKS_API_KEY local-dev validation
- server.py: provider-aware fallback kwargs build
- sanitize_thinking_blocks: also sanitize ChatBedrockConverse thinking blocks
- model_fallback: treat transient botocore ClientError codes as fallback-worthy
- eval_jobs: repoint hardcoded eval model id to Bedrock Claude
- tests: repoint dropped model ids; drop obsolete google test module

* fix(bedrock): use adaptive thinking + output_config.effort for Opus 4.8

The handoff spec wired Bedrock Converse thinking as
{type: enabled, budget_tokens: N}, but Opus 4.7+ rejects that with a
ValidationException: thinking.type "enabled" is not supported; it requires
thinking.type "adaptive" plus output_config.effort. Verified by live invoke
against us.anthropic.claude-opus-4-8 (account 328440206208, us-east-1):
the enabled+budget shape 400s, adaptive+effort returns normally.

Map profile effort to additional_model_request_fields:
  {thinking: {type: adaptive, display: summarized},
   output_config: {effort: <low|medium|high|xhigh|max>}}
reusing anthropic_thinking_for/anthropic_effort_for. Update the two
subagent-model tests asserting the old shape.

* fix(deploy): seed Bedrock/Fireworks models, not the dropped anthropic:/openai: ids

Model selection is store-driven, so seed_store.sh's team_settings/default seed is
what runs in prod. It still seeded the removed providers, which would fail at runtime
after the migration:
- agent/builder: anthropic:claude-opus-4-8 -> bedrock_converse:us.anthropic.claude-opus-4-8
- reviewer: openai:gpt-5.5 (dropped) -> bedrock_converse:us.anthropic.claude-opus-4-8
  (set SEED_REVIEWER_MODEL to a Fireworks model for a cross-family reviewer)
- fetch-config REQUIRED_PROVIDER_KEYS default ANTHROPIC_API_KEY,OPENAI_API_KEY ->
  FIREWORKS_API_KEY (Bedrock auths via host IAM role; dropping the old keys would
  otherwise fail-fast at boot)
- docs (DEPLOYMENT/ROTATION/put-config) updated to match.

Surfaced by the cross-family review + verified against deploy/.

* fix(bedrock): security-review NITs — region resolution, error sanitization, reasoning-block strip

From /sh-security-review (all confirmed-low):
- model.py: resolve region from AWS_REGION OR AWS_DEFAULT_REGION (matches
  validate_local_dev_llm_config) so the validated region is the one actually used.
- model_fallback.py: sanitize Bedrock AccessDenied/ResourceNotFound errors to the
  error code only, so the role ARN + account id in the raw botocore message never
  reach logs or the user channel (CWE-209).
- sanitize_thinking_blocks.py: also strip empty Bedrock reasoning_content blocks
  (Converse emits reasoning_content, not thinking) so the middleware is not a no-op
  on Bedrock; + unit tests. (Empty blocks replay fine today; defensive.)

* deploy(bedrock): grant instance-role Bedrock invoke + repoint LLM_MODEL_ID / eval model ids

Deployment-readiness for the Bedrock migration (PR #62):
- instance-role.ts: least-privilege bedrock:InvokeModel[WithResponseStream] on the
  us.anthropic.claude-opus-4-8 inference-profile ARN + the foundation-model ARN in
  each routed region (us-east-1/2, us-west-2). The model runs in the server process
  on the box, so the EC2 instance role is the principal. Simulator-verified (allowed
  for opus-4-8, implicitDeny for other models) and synth-verified. Passed the
  mandatory GPT-4.1 IAM cross-review (no blockers, least-privilege confirmed).
- config-store.ts: IaC SSM LLM_MODEL_ID anthropic:claude-opus-4-8 ->
  bedrock_converse:us.anthropic.claude-opus-4-8. This SSM value overrides
  seed_store.sh's default via pick precedence, so the seed-script fix alone was
  insufficient — both sources now point at the supported Bedrock id.
- infra/README.md + evals/reviewer/config.toml: repoint stale anthropic:/google_genai:
  ids to the Bedrock id (config.toml's model_id was an active, now-broken value).

AWS_REGION is already wired via user-data.sh (IMDS -> boot.env), so no change needed there.

* chore(secrets): drop OPENAI/GOOGLE/GROQ key shells (revoked, providers removed)

Those three providers were dropped in the Bedrock/Fireworks migration and their keys
revoked; the live Secrets Manager objects (open-swe-{dev,prod}/{OPENAI,GOOGLE,GROQ}_API_KEY)
were deleted (7-day recovery). Remove them from the IaC so a future cdk deploy does not
recreate the shells, and from fetch-config's mirror array so boot stops requesting them:
- config-store.ts SECRET_VARS + descriptions (28 -> 25 shells)
- fetch-config.sh SECRET_VARS array (kept in lockstep)
- put-config.sh: drop the put_secret lines; ANTHROPIC_API_KEY re-labelled optional
  (eval judge only — Bedrock builder/reviewer auth via the host IAM role).

REQUIRED_PROVIDER_KEYS is not set in SSM, so it uses the FIREWORKS_API_KEY default.
2026-06-29 15:57:19 -04:00

152 lines
8.2 KiB
Bash
Executable file

#!/usr/bin/env bash
# put-config.sh — out-of-band populator for the open-swe config store.
#
# T11 (CDK) creates the RESOURCE SHELLS:
# - 28 Secrets Manager secrets open-swe-<env>/<VAR> (value-LESS shells)
# - the IaC-managed SSM params /open-swe-<env>/<VAR> (real values, owned in CDK)
# This script sets the values that CANNOT live in IaC — every secret value, plus
# the out-of-band SSM params (operationally-variable / env-specific-unknown). Run
# it AFTER `cdk deploy open-swe-<env>` and BEFORE the EC2/T12 box first boots, so
# fetch-config.sh finds every REQUIRED var populated and never writes a partial .env.
#
# The secret list below mirrors config-store.ts SECRET_VARS (28 names — the
# inventory's "29" double-counted JUDGE_ANTHROPIC_BASE_URL, which is config
# (SSM), not a secret). Keep the two lists in lockstep.
#
# SAFETY:
# - NO real secret values live in this file — every secret is a <FILL> placeholder.
# Replace <FILL...> inline at run time, pipe from a vault, or export the
# matching OPENSWE_PUT_<VAR> env var; NEVER commit real values.
# - It does NOT touch the IaC-managed SSM params (SANDBOX_TYPE, DEFAULT_REPO_OWNER,
# ALLOWED_GITHUB_ORGS, DEFAULT_REPO_NAME, DASHBOARD_*_URL/ORIGINS, LLM_MODEL_ID)
# — CDK owns those; setting them here would cause drift.
# - Secrets go to Secrets Manager; the box's instance role grants read on
# open-swe-<env>/* and /open-swe-<env>/* (no kms:Decrypt — AWS-managed keys).
#
# Idempotent: put-secret-value adds a new AWSCURRENT version; put-parameter
# --overwrite updates in place.
set -euo pipefail
ENV="${1:-}"
case "$ENV" in
dev | prod) ;;
*)
echo "usage: put-config.sh <dev|prod>" >&2
exit 2
;;
esac
REGION="${AWS_REGION:-${AWS_DEFAULT_REGION:-us-east-1}}"
SECRET_PREFIX="open-swe-${ENV}/"
SSM_PREFIX="/open-swe-${ENV}/"
command -v aws >/dev/null 2>&1 || { echo "put-config: 'aws' not found on PATH" >&2; exit 3; }
# --- helpers -----------------------------------------------------------------
# put_secret VAR : set the value of an EXISTING secret shell open-swe-<env>/VAR.
# Resolution order for the value: $OPENSWE_PUT_<VAR> env var, else the literal
# <FILL> placeholder (which aborts so an unset secret is never silently shipped).
put_secret() {
local var="$1"
local override_name="OPENSWE_PUT_${var}"
local value="${!override_name:-<FILL>}"
if [ "$value" = "<FILL>" ]; then
echo "put-config[$ENV]: SKIP secret ${var} (no value; set ${override_name} or edit inline)" >&2
return 0
fi
aws secretsmanager put-secret-value \
--secret-id "${SECRET_PREFIX}${var}" \
--secret-string "$value" \
--region "$REGION" \
--no-cli-pager >/dev/null
echo "put-config[$ENV]: set secret ${var}"
}
# put_param VAR [TYPE] : create/update an out-of-band SSM param /open-swe-<env>/VAR.
# TYPE defaults to String; pass SecureString for anything sensitive-but-not-a-secret.
put_param() {
local var="$1" type="${2:-String}"
local override_name="OPENSWE_PUT_${var}"
local value="${!override_name:-<FILL>}"
if [ "$value" = "<FILL>" ]; then
echo "put-config[$ENV]: SKIP param ${var} (no value; set ${override_name} or edit inline)" >&2
return 0
fi
aws ssm put-parameter \
--name "${SSM_PREFIX}${var}" \
--value "$value" \
--type "$type" \
--overwrite \
--region "$REGION" \
--no-cli-pager >/dev/null
echo "put-config[$ENV]: set param ${var} (${type})"
}
# --- 1) Secrets (open-swe-<env>/<VAR>) — 28 shells from config-store.ts -------
# REQUIRED at boot (fetch-config fail-fast): DASHBOARD_JWT_SECRET,
# TOKEN_ENCRYPTION_KEY, GITHUB_APP_PRIVATE_KEY/CLIENT_SECRET, the active provider
# key(s) (ANTHROPIC_API_KEY + OPENAI_API_KEY by default), LANGSMITH_API_KEY_PROD,
# and (prod) GITHUB_WEBHOOK_SECRET + SLACK_SIGNING_SECRET.
put_secret ANTHROPIC_API_KEY # optional — eval judge only (JUDGE_ANTHROPIC_API_KEY fallback); Bedrock builder/reviewer use the host IAM role
put_secret DASHBOARD_JWT_SECRET # REQUIRED — dashboard session JWT signing
put_secret TOKEN_ENCRYPTION_KEY # REQUIRED — Fernet key(s) for GH-token crypto
put_secret GITHUB_APP_PRIVATE_KEY # REQUIRED — GitHub App PEM (multiline; quote it)
put_secret GITHUB_APP_CLIENT_SECRET # REQUIRED — dashboard OAuth login
put_secret LANGSMITH_API_KEY_PROD # REQUIRED (langsmith sandbox) — prod key
put_secret GITHUB_WEBHOOK_SECRET # prod-REQUIRED — GitHub webhook signature
put_secret SLACK_SIGNING_SECRET # prod-REQUIRED — Slack webhook signature
put_secret LINEAR_WEBHOOK_SECRET # required only when Linear is wired
# Optional / conditional secrets — set the ones this deployment actually uses.
put_secret CORRIDOR_API_TOKEN # optional — Corridor MCP
put_secret CORRIDOR_MCP_TOKEN # optional — Corridor MCP (alt name)
put_secret CORRIDOR_TOKEN # optional — Corridor MCP (alt name)
put_secret DAYTONA_API_KEY # only if SANDBOX_TYPE=daytona
put_secret EXA_API_KEY # optional — Exa web search
put_secret FIREWORKS_API_KEY # active non-Claude key (fallback/subagents); Bedrock uses the host IAM role
put_secret GITHUB_PAT # optional — PAT fallback
put_secret JUDGE_ANTHROPIC_API_KEY # optional — eval judge (falls back to ANTHROPIC)
put_secret LANGSMITH_API_KEY # optional — LangSmith (dev)
put_secret LANGCHAIN_API_KEY # optional — LangSmith alt name
put_secret LINEAR_API_KEY # optional — Linear API
put_secret RUNLOOP_API_KEY # only if SANDBOX_TYPE=runloop
put_secret SLACK_BOT_TOKEN # optional — Slack bot token
put_secret SLACK_CLIENT_SECRET # optional — Slack OAuth
put_secret USER_ID_API_KEY_MAP # optional — JSON map user id -> API key (per-user auth)
put_secret X_SERVICE_AUTH_JWT_SECRET # optional — service-auth JWT
# --- 2) Out-of-band SSM params (/open-swe-<env>/<VAR>) ------------------------
# These are NOT created by CDK (operationally-variable / env-specific-unknown).
# put_param creates them on first run.
put_param DEFAULT_SANDBOX_SNAPSHOT_ID # REQUIRED (langsmith) — changes every rebuild
put_param GITHUB_APP_ID # REQUIRED — GitHub App numeric id
put_param GITHUB_APP_INSTALLATION_ID # REQUIRED — GitHub App installation id
put_param GITHUB_APP_CLIENT_ID # REQUIRED — dashboard OAuth client id
put_param GITHUB_OAUTH_PROVIDER_ID # GitHub OAuth provider id
put_param LANGSMITH_TENANT_ID_PROD # LangSmith prod tenant id
put_param LANGSMITH_URL_PROD # LangSmith prod URL
put_param LANGSMITH_ENDPOINT # LangSmith API endpoint
put_param LANGSMITH_ENDPOINT_PROD # LangSmith prod API endpoint
put_param LANGSMITH_HOST_API_URL # LangSmith host API URL
put_param LANGGRAPH_URL # LangGraph server URL
put_param LANGGRAPH_URL_PROD # LangGraph server URL (prod)
put_param LANGCHAIN_REVISION_ID # LangChain revision id
put_param SLACK_CLIENT_ID # Slack OAuth client id
put_param SLACK_TEAM_ID # Slack workspace/team id
put_param SLACK_BOT_USER_ID # Slack bot user id
put_param SLACK_BOT_USERNAME # Slack bot username
put_param SLACK_REPO_OWNER # Slack default repo owner
put_param SLACK_REPO_NAME # Slack default repo name
put_param CONFIGURED_ADMINS # dashboard admin GitHub logins (comma list)
put_param OBSERVABILITY_AUTHORIZED_EMAILS # observability allowlist (comma list)
put_param PUBLIC_REPO_ORG_GATE # public-repo trigger gate (org name; empty=off)
put_param ALLOWED_GITHUB_REPOS # extra repo allowlist (comma list)
put_param LLM_FALLBACK_MODEL_ID # optional — model fallback id
put_param DATADOG_MCP_TOOLSETS # optional — Datadog MCP toolsets
put_param NOTION_MCP_CLIENT_NAME # optional — Notion MCP client name
put_param API_STANDARDS_SKILL_HANDLE # optional — API standards skill handle
put_param REPO_SNAPSHOT_BASE_IMAGE # optional — repo snapshot base image
put_param REPO_SNAPSHOT_BUILD_TIMEOUT_SECONDS # optional — snapshot build timeout
put_param REPO_SNAPSHOT_STALE_BUILD_SECONDS # optional — snapshot stale threshold
echo "put-config[$ENV]: done. Verify with fetch-config.sh ${ENV} before first boot."