feat(quoter-bot): deploy one railway service per chain

[haydenshively]

BOTS-63. The mainnet code path landed in #173; this is the deployment half: the quoter-bot Railway
project now runs one quoter-bot-<chainId> service per supported chain (quoter-bot-1,
quoter-bot-8453), the way the liquidation bots run bot-<chainId> (#193).

Shape

Unlike the liquidation bots, per-chain inputs are not _<chainId>-suffixed. The quoter has
about a dozen chain-specific inputs including two large JSON strategy arrays, so a full run keeps
every input under its runtime name and provisions the one chain named by CHAIN_ID. Run it once per
chain with that chain's environment.

• Full run: CHAIN_ID is parsed with the bot's own chainIdValue, so the script rejects exactly
  what the runtime rejects (0x1, 1.0, unsupported ids) before touching Railway. It then
  provisions quoter-bot-<chainId>: service, /state volume, Dockerfile path, variables.
• DEPLOY_ONLY (CI): the target set is SUPPORTED_CHAIN_IDS.map(railwayServiceName), so the
  deploy set cannot drift from the bot's chain set. It preflights that every service exists and
  names the missing ones (deploy-only holds no secrets and cannot create), then re-ships through
  reshipRailwayServices: every upload starts before any wait begins, a rejected start is recorded
  as START_FAILED and a failed status read as POLL_FAILED, and the run exits non-zero unless
  every service reaches SUCCESS. One chain's failure never blocks another's re-ship, which matters
  because mainnet is expected to halt at readiness until its market is listed.

No workflow job change: deploy-quoter-bot-production.yml already runs the script with
DEPLOY_ONLY=true and the Docker Hub / npm jobs hang off that single job. Only its header comment
changed. README and docs/reference.md deploy sections updated; the stale chainId: 8453 log claim
is now "the configured chainId".

Cutover (manual, not committed)

1. The live quoter-bot service has been renamed to quoter-bot-8453: the service id is
   unchanged, and railway variable list still shows RAILWAY_VOLUME_ID / RAILWAY_VOLUME_MOUNT_PATH,
   so the /state volume and all variables survived. A DEPLOY_ONLY preflight against production
   from this branch now reports exactly: Not provisioned in production: quoter-bot-1.
2. quoter-bot-1 still needs a full run with mainnet config (CHAIN_ID=1, mainnet RPC + archive
   RPC, mainnet MIDNIGHT_ADDRESS / RATIFIER_ADDRESS / USDC / MARKET_IDS / strategy arrays, its
   own funded maker with an unbounded USDC approval, a second BetterStack heartbeat URL).
3. Merge without the release-quoter-bot label until step 2 is done: a labelled merge deploys
   immediately and DEPLOY_ONLY fails while quoter-bot-1 is unprovisioned.

Blocker outside this repo

No mainnet Midnight market is listed=true yet (/v0/midnight/markets?chain_ids=1&listed=true
returns []). getBook maps listed to allowlisted, setup-check reports not allowlisted, and a
non-transient readiness failure halts the monitor, so quoter-bot-1 will exit at startup until the
market is listed. Provisioning ahead of listing is safe; calling the release done is not.

Verification

• pnpm --filter @morpho-org/quoter-bot run typecheck, jsdoc:check, root pnpm lint,
  pnpm format, pnpm knip clean.
• Root pnpm test: 3030 pass; the only failures are the four fork suites that need RPC_URL_8453
  (CI has it). test/scripts/railway.utils.test.ts: 23 tests, including a runnable
  reshipRailwayServices test where one start rejects, another's poll fails, the third succeeds,
  and the aggregate is non-zero.
• Codex (GPT Sol, high) reviewed the plan and the implementation; all findings incorporated except
  one deferred: binding the deploy poll to the railway up --json deployment id is pre-existing and
  serialized today by the workflow's concurrency group.

🤖 Generated with Claude Code

[linear-code]

BOTS-63

BOTS-106

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-07T06:08:02.480261Z	4f6aa13	Draft marked ready

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[devin-ai-integration]

✅ Devin Review: No Issues Found

Devin Review analyzed this PR and found no bugs or issues to report.

[haydenshively]

@julien-devatom regarding cutover step 2 -- can you handle this, or let me know who I should coordinate with to get the proper config?

[julien-devatom]

@codex review

[julien-devatom]

@julien-devatom regarding cutover step 2 -- can you handle this, or let me know who I should coordinate with to get the proper config?

Will do this morning

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. Delightful!

Reviewed commit: 4f6aa133aa

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".