feat: self-hosted OpenAI-compatible providers (LiteLLM) for Pi#416

Open

hauserkristof wants to merge 22 commits into

getsentry:mainfrom

hauserkristof:feat/custom-pi-provider-litellm

Commits on Jun 26, 2026

docs: design for self-hosted LLMs via custom Pi providers (LiteLLM)

hauserkristof
and
claude
committed
docs: implementation plan for custom Pi providers (LiteLLM)

hauserkristof
and
claude
committed
feat(config): add custom provider schema for self-hosted LLMs

hauserkristof
and
claude
committed
feat(runtime): normalize custom providers and resolve keys from env

hauserkristof
and
claude
committed
fix(runtime): drop unreachable IPv6 branch, dedupe provider-name sanitize

hauserkristof
and
claude
committed
fix(runtime): keep bracketed IPv6 loopback branch (URL.hostname brackets ::1)

hauserkristof
and
claude
committed
feat(runtime): expose custom providers via getRuntimeProviderOptions

hauserkristof
and
claude
committed
feat(runtime): register custom providers in the pi adapter

hauserkristof
and
claude
committed
feat(runtime): thread custom providers through all model lanes

hauserkristof
and
claude
committed
feat(config): propagate custom providers to all runner entry points

hauserkristof
and
claude
committed
feat(skill-builder): route build/improve through custom providers

hauserkristof
and
claude
committed
feat(runtime): fail fast when a remote custom provider has no key

hauserkristof
and
claude
committed
test(cli): make custom-provider preflight tests hermetic

hauserkristof
and
claude
committed
docs: document self-hosted OpenAI-compatible providers (LiteLLM)

hauserkristof
and
claude
committed
docs,runtime: clarify env-resolution boundary and IPv6 loopback

hauserkristof
and
claude
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

feat: self-hosted OpenAI-compatible providers (LiteLLM) for Pi#416

feat: self-hosted OpenAI-compatible providers (LiteLLM) for Pi#416
hauserkristof wants to merge 22 commits into
getsentry:mainfrom
hauserkristof:feat/custom-pi-provider-litellm

Commits on Jun 26, 2026

docs: design for self-hosted LLMs via custom Pi providers (LiteLLM)

docs: implementation plan for custom Pi providers (LiteLLM)

feat(config): add custom provider schema for self-hosted LLMs

feat(runtime): normalize custom providers and resolve keys from env

fix(runtime): drop unreachable IPv6 branch, dedupe provider-name sanitize

fix(runtime): keep bracketed IPv6 loopback branch (URL.hostname brackets ::1)

feat(runtime): expose custom providers via getRuntimeProviderOptions

feat(runtime): register custom providers in the pi adapter

feat(runtime): thread custom providers through all model lanes

feat(config): propagate custom providers to all runner entry points

feat(skill-builder): route build/improve through custom providers

feat(runtime): fail fast when a remote custom provider has no key

test(cli): make custom-provider preflight tests hermetic

docs: document self-hosted OpenAI-compatible providers (LiteLLM)

docs,runtime: clarify env-resolution boundary and IPv6 loopback

Commits on Jun 30, 2026

fix(runtime): apply custom-provider preflight consistently across entry points

fix(runtime): inherit the default model for auxiliary and synthesis lanes

fix(action): inherit the default model for the PR workflow auxiliary lane

fix(action): align runtime and provider-options resolution in fix evaluation

fix(config): inherit base-layer custom providers into the repo layer

fix(cli): bring warden build/improve in line for custom providers

fix(skill-builder): forward custom providers to the fallback JSON repair