Code AI platform with Code Search & Cody
Go to file
Taras Yemets d19aa106f9
feat(cody): add circuit breaker to handle timed-out requests and rate limit hits (#64133)
<!-- PR description tips:
https://www.notion.so/sourcegraph/Write-a-good-pull-request-description-610a7fd3e613496eb76f450db5a49b6e
-->
Closes
https://linear.app/sourcegraph/issue/CODY-2758/[autocomplete-latency]-add-circuit-breaker-in-cody-gateway-to-handle

This PR introduces an in-memory model availability tracker to ensure we
do not send consequent requests to the currently unavailable LLM
providers.

The tracker maintains a history of error records for each model. It
evaluates these records to determine whether the model is available for
new requests. The evaluation follows these steps:
1. For every request to an upstream provider, the tracker records any
errors that occur. Specifically, it logs timeout errors (when a request
exceeds its deadline) and responses with a 429 status code (Too Many
Requests).
2. These error records are stored in a circular buffer for each model.
This buffer holds a fixed number of records, ensuring efficient memory
usage.
3. The tracker calculates the failure ratio by analyzing the stored
records. It checks the percentage of errors within a specified
evaluation window and compares this against the total number of recent
requests.
4. Based on the calculated failure ratio, the tracker decides whether
the model is available:
- Model Unavailable: If the ratio of failures (timeouts or 429 status
codes) exceeds a predefined threshold (X%), the model is marked as
unavailable. In this state, the system does not send new requests to the
upstream provider.
When a model is unavailable, the system immediately returns an error
status code, typically a 503 Service Unavailable, to the client. This
informs the client that the service is temporarily unavailable due to
upstream issues.
- Model Available: If the failure ratio is within acceptable limits, the
system proceeds with sending the request to the upstream provider.

This PR suggests considering a model unavailable if **95% of the last
100 requests within the past minute** either time out or return a 429
status code. I am not sure about these exact values and suggest them as
a starting point for discussion

## Test plan
- Added unit tests
- CI
<!-- REQUIRED; info at
https://docs-legacy.sourcegraph.com/dev/background-information/testing_principles
-->

## Changelog

<!-- OPTIONAL; info at
https://www.notion.so/sourcegraph/Writing-a-changelog-entry-dd997f411d524caabf0d8d38a24a878c
-->
2024-08-05 11:38:58 +00:00
.apko Build images end-to-end using Bazel v2 (#61845) 2024-04-12 16:18:43 +01:00
.aspect chore(bazel): re-enable remote caching of Copy{File,Directory,ToDirectory} actions (#64153) 2024-07-30 16:02:11 +00:00
.buildkite chore(ci): remove Percy visual tests (#63515) 2024-06-27 16:20:06 +02:00
.github Revert "chore(ci): post instructions for PR sections in a comment" (#64187) 2024-07-31 13:10:06 +01:00
.vscode feat(search): Make search aware of perforce changelist id mapping (#63563) 2024-07-09 14:01:05 -04:00
client [Svelte]: UI Updates for Perforce Depots and Git Repos (#64014) 2024-08-02 20:47:20 -06:00
cmd feat(cody): add circuit breaker to handle timed-out requests and rate limit hits (#64133) 2024-08-05 11:38:58 +00:00
dev Enable p4-fusion by default for Perforce code host connections (#64101) 2024-08-02 11:15:16 +02:00
doc fix(docs): update quickstart guide links (#64202) 2024-08-01 16:18:55 -07:00
docker-images feat(search): Enable improved symbol parsing for large repos (when using Rockskip) (#63988) 2024-07-31 15:27:31 -04:00
internal chore: Remove incorrectly logged warning (#64267) 2024-08-05 09:43:43 +00:00
lib servicecatalog: remove searcher dep on database (#64247) 2024-08-05 09:55:49 +02:00
migrations chore(search_jobs): add janitor job (#64186) 2024-08-01 15:29:10 +02:00
monitoring feat(search): remove smart search logic (#64215) 2024-08-01 18:02:35 +03:00
schema Enable p4-fusion by default for Perforce code host connections (#64101) 2024-08-02 11:15:16 +02:00
testing feat/bazel: //cmd/{frontend,server} targets that don't include client bundle for backend integration tests (#62877) 2024-05-28 14:32:48 +01:00
third_party enterprise-portal: implement basic MSP IAM and RPCs (#63173) 2024-06-19 21:46:48 -04:00
third-party-licenses finish removing chromatic (#63966) 2024-07-21 18:37:02 -07:00
tools Integrate security release approval into release pipeline (#63990) 2024-07-24 09:19:49 +01:00
ui/assets feat(frontend): do not embed frontend assets anymore (#63946) 2024-07-31 15:17:52 +02:00
wolfi-images security: Auto-update package lockfiles for Sourcegraph base images (#64204) 2024-08-02 17:54:55 +02:00
wolfi-packages Unpin openjdk-11 version in blobstore package (#64045) 2024-07-24 16:57:21 +01:00
.bazel_fix_commands.json SG Start Bazel Improvements Take 2 (#60687) 2024-03-05 01:44:21 -08:00
.bazelignore Convert Appliance Maintenance UI to Bazel (#63661) 2024-07-10 13:47:18 +02:00
.bazeliskrc chore: upgrade to Aspect CLI 5.8.19 (#59203) 2024-01-02 15:13:24 +01:00
.bazelrc feat(ci): Adds playwright tests for sveltekit to bazel (#62560) 2024-06-06 12:45:05 -06:00
.bazelversion chore(bazel): bump to 7.2.0 (#63226) 2024-06-12 13:25:18 +00:00
.dockerignore use esbuild for client/web builds (#57365) 2023-10-23 10:59:06 -07:00
.editorconfig proto: Add editorconfig to ident using two spaces (#57281) 2023-10-03 00:39:42 +00:00
.eslintrc.js finish removing chromatic (#63966) 2024-07-21 18:37:02 -07:00
.gitattributes dev/linearhooks: add POC (#62367) 2024-05-07 00:14:05 -07:00
.gitignore chore(ci): emit compact executon log in CI (#63420) 2024-06-21 19:50:35 +01:00
.graphqlrc.yml
.hadolint.yaml bump comby version to 1.7.1 (#35830) 2022-05-20 20:12:01 -07:00
.mailmap mailmap: add entries for Eric and Renovate (#50966) 2023-04-25 09:42:22 +02:00
.mocharc.js reapply "switch from jest to vitest for faster, simpler tests (#57886)" (#58145) 2023-11-07 12:00:18 +02:00
.npmrc pnpm: remove update notifier message (#51630) 2023-05-10 08:53:39 +02:00
.pre-commit-config.yaml chore(local): add FORBIDCOMMIT pragma to prevent accidental commits (#63581) 2024-07-01 18:27:26 +00:00
.prettierignore feat/dotcom: use Enterprise Portal for Cody Gateway usage (#63653) 2024-07-10 19:22:08 +00:00
.stylelintignore rework plugin structure and implement frontside blogpost (#46883) 2023-02-15 11:49:51 +02:00
.stylelintrc.json web: drop bootstrap depenedency (#41401) 2022-09-07 03:11:26 -07:00
.swcrc use swc instead of babel for faster bazel typescript transpilation (#57912) 2023-11-02 22:49:03 -07:00
.tool-versions chore(tooling): bump Go version to 1.22.4 (#63124) 2024-06-06 15:19:03 +00:00
.trivyignore
BUILD.bazel feat(frontend): do not embed frontend assets anymore (#63946) 2024-07-31 15:17:52 +02:00
CHANGELOG.md feat(code insights): language stats speed improvements by using archive loading (#62946) 2024-07-18 08:40:48 +02:00
CODENOTIFY nix: update pnpm hash (#51512) 2023-05-05 12:51:59 +00:00
CONTRIBUTING.md fix: update links for dev docs (#62758) 2024-05-17 13:47:34 +02:00
deps.bzl gomod: update zoekt for shard scanning improvement (#64264) 2024-08-05 12:36:27 +02:00
doc.go
eslint-relative-formatter.js bazel: implement custom ESLint Bazel rule (#52062) 2023-05-22 04:05:45 -07:00
flake.lock nix: bump to bazel 7.1 (#61326) 2024-03-22 16:57:50 +00:00
flake.nix nix: use go1.22.4 (#63372) 2024-06-20 11:12:17 +02:00
gen.go chore: fixup go-mockgen run statement (#61028) 2024-03-12 13:06:36 +00:00
go.mod gomod: update zoekt for shard scanning improvement (#64264) 2024-08-05 12:36:27 +02:00
go.sum gomod: update zoekt for shard scanning improvement (#64264) 2024-08-05 12:36:27 +02:00
graphql-schema-linter.config.js
LICENSE relicense all paths other than MIT licensed code, client/cody*, jetbrains, VS code, and browser extension to enterprise (#53345) (#53345) 2023-06-13 10:28:11 -07:00
LICENSE.enterprise Update Enterprise license copyright notice (#62467) 2024-05-06 17:35:32 +00:00
linter_deps.bzl feat(build): Add exhaustruct linter (codeintel POC) (#63965) 2024-07-23 17:26:01 +08:00
mockgen.temp.yaml chore(codeintel): Rename uploads/../Store -> codegraph/../DataStore (#64001) 2024-07-24 08:24:20 +01:00
mockgen.test.yaml chore: Move authn into cmd/frontend (#63648) 2024-07-31 03:26:25 +02:00
mockgen.yaml bazel: native go-mockgen in Bazel (#60386) 2024-02-16 13:26:48 +00:00
nogo_config.json chore: Remove redundant loop captures (#62264) 2024-04-30 07:57:21 -06:00
package.json Cody Web: Update Cody Web to 0.3.6 [React version] (#64254) 2024-08-02 18:04:52 -03:00
pnpm-lock.yaml Cody Web: Update Cody Web to 0.3.6 [React version] (#64254) 2024-08-02 18:04:52 -03:00
pnpm-workspace.yaml Convert Appliance Maintenance UI to Bazel (#63661) 2024-07-10 13:47:18 +02:00
postcss.config.js
prettier.config.js clean up Cody CSS to increase shareability and improve display in web app (#50279) 2023-04-03 12:29:05 -07:00
README.md chore: remove broken link in README (#63256) 2024-06-13 22:22:56 +00:00
release.yaml fix(release): check for marker in container files instead of binary (#64205) 2024-08-01 08:25:30 +00:00
renovate.json chore(ci): disable renovate (#63313) 2024-06-19 13:17:15 +02:00
SECURITY.md
service-catalog.yaml lib/servicecatalog: init to distribute catalog (#46999) 2023-01-26 17:22:27 -08:00
sg.config.yaml fix(sg): acknowledge command execution state to avoid recursion when executing short running commands (#64181) 2024-07-31 16:09:44 -05:00
shell.nix bazel: use pgutil binaries from GCS instead of from the host (#61741) 2024-04-11 18:00:21 +01:00
stamp_tags.bzl Switch to OCI/Wolfi based image (#52693) 2023-06-02 12:12:52 +02:00
tsconfig.base.json web: fix pnpm-lock issue (#47478) 2023-02-09 22:04:31 -08:00
tsconfig.json release: drop legacy release tooling (#61220) 2024-04-09 14:29:35 -05:00
vitest.shared.ts make pagination hooks store filter & query params in URL, not just pagination params (#63744) 2024-07-15 19:17:59 +00:00
vitest.workspace.ts vitest: Fix workspace config wrt client/web/ (#58397) 2023-11-17 08:22:46 +00:00
WORKSPACE release/bug: generate a new stitched migration graph (#63764) 2024-07-10 14:49:18 -07:00

DocsContributingTwitterDiscord

Build status Scorecard Latest release Discord Contributors


Sourcegraph makes it easy to read, write, and fix code—even in big, complex codebases.

  • Code search: Search all of your repositories across all branches and all code hosts.
  • Code intelligence: Navigate code, find references, see code owners, trace history, and more.
  • Fix and refactor: Roll out large-scale changes to many repositories at once and track big migrations.

Getting started



Development

Refer to the Developing Sourcegraph guide to get started.

Documentation

The doc directory has additional documentation for developing and understanding Sourcegraph:

License

This repository contains primarily non-OSS-licensed files. See LICENSE.

Copyright (c) 2018-present Sourcegraph Inc.