fix(helm): Enable Backend Deployment to mount Gateway config.yaml

[tin-berri]

Problem

• Backend Deployment never mounted config.yaml or set CONFIG_FILE_PATH; only the gateway did
• So cache, callbacks, and guardrails from gateway.config.proxy_config only applied to the gateway, never the backend
• Backend runs the same proxy code and needs the same settings, so things like a shared cache silently didn't work on backend pods leading to cross-pod sync failures when adding MCP servers

Fix

• Mount the gateway's existing ConfigMap on the backend, set CONFIG_FILE_PATH, add a checksum/config annotation so backend pods roll on config changes
• Gated on the existing gateway.config.create; when false, backend is unchanged
• Note: Reason to mirror gateway's existing ConfigMap is to keep the gateway proxy config as our single source of truth. This limits the risk of litellm-gateway and litellm-backend to have separate proxy config and breaking shared settings for whichever side is forgotten/behind on changes.

Linear ticket

Resolves LIT-3466
Resolves LIT-3246 (1st Problem)
Resolves LIT-3547

Tests

• Test that endpoints from the litellm-backend bod that use to rely on in-memory memory is able to have cross-pod performance via the shared redis cache after we mounted a cache configuration in the CONFIG_FILE_PATH
• Used helm template to verify that togglinggateway.config.create=true allows backend deployment to contain the CONFIG_FILE_PATH

Pre-Submission checklist

Please complete all items before asking a LiteLLM maintainer to review your PR

• I have added meaningful tests
• My PR passes all unit tests on make test-unit
• My PR's scope is as isolated as possible; it only solves 1 specific problem
• I have requested a Greptile review by commenting @greptileai and received a Confidence Score of at least 4/5 before requesting a maintainer review

Delays in PR merge?

If you're seeing a delay in your PR being merged, ping the LiteLLM Team on Slack (#pr-review).

CI (LiteLLM team)

CI status guideline:

• 50-55 passing tests: main is stable with minor issues.
• 45-49 passing tests: acceptable but needs attention
• <= 40 passing tests: unstable; be careful with your merges and assess the risk.

• Branch creation CI run
  Link:

• CI run for the last commit
  Link:

• Merge / cherry-pick CI run
  Links:

Screenshots / Proof of Fix

Type

🐛 Bug Fix

Changes

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you all sign our Contributor License Agreement before we can accept your contribution.
1 out of 3 committers have signed the CLA.

✅ yuneng-berri
❌ Tin Chi Lo
❌ shin-berri

---

Tin Chi Lo seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[tin-berri]

@greptileai

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[greptile-apps]

Greptile Summary

This PR extends the Helm chart's existing gateway config-from-ConfigMap pattern to the backend component, allowing users to supply a proxy_config that gets mounted as /app/config/config.yaml and wired via the CONFIG_FILE_PATH env var.

• Adds helm/litellm/templates/backend/configmap.yaml — a new template identical in structure to the gateway's configmap, gated on backend.config.create.
• Updates helm/litellm/templates/backend/deployment.yaml to conditionally add the checksum annotation, CONFIG_FILE_PATH env var, volumeMount, and pod-level volumes when backend.config.create: true.
• Adds backend.config: {create: false, proxy_config: {}} to values.yaml, defaulting the feature to off so existing deployments are unaffected.

Confidence Score: 5/5

Safe to merge — the feature is opt-in (defaults to disabled) and faithfully replicates the already-deployed gateway config pattern without touching any existing code paths.

All three changes are additive and gated behind backend.config.create: false by default, so no existing deployment is affected. The implementation is a near-exact copy of the gateway config wiring that is already in production, reducing the risk of novel errors.

No files require special attention.

Important Files Changed

Filename	Overview
helm/litellm/templates/backend/configmap.yaml	New ConfigMap template for backend proxy config, exact mirror of the gateway configmap pattern.
helm/litellm/templates/backend/deployment.yaml	Adds checksum annotation, CONFIG_FILE_PATH env var, volumeMount, and volumes section for the backend config — mirrors the gateway deployment closely.
helm/litellm/values.yaml	Adds config: {create: false, proxy_config: {}} under backend, matching the existing gateway config shape. Defaults to disabled.

_{Reviews (2): Last reviewed commit: "omit backend annotations block when conf..." | Re-trigger Greptile}

[mateo-berri]

From Yassin: please fix greptile and merge

[mateo-berri]

LGTM; thanks!