feat(evaluators): DO NOT MERGE Test end to end#221
Open
namrataghadi-galileo wants to merge 46 commits into
Open
feat(evaluators): DO NOT MERGE Test end to end#221namrataghadi-galileo wants to merge 46 commits into
namrataghadi-galileo wants to merge 46 commits into
Conversation
Mirrors #204's bindings migration: replaces require_admin_key and router-level require_api_key with require_operation(CONTROLS_*) on every protected route on /controls and on /control-templates/render. Both routers now mount with the non-validating get_api_key_from_header so the framework owns authentication and authorization, with the extractor attached purely so the generated OpenAPI advertises X-API-Key. GET /controls/schema is intentionally left without a require_operation dependency: it returns a static model schema with no tenant state and routing it through the framework would force the upstream provider to handle a meta-only operation that has no permission semantics. POST /controls/validate and POST /control-templates/render are wired to CONTROLS_CREATE rather than CONTROLS_READ. Both exercise the authoring materialization path and exist to support the create / set- data flow; a caller who cannot create controls has no use for the result. Backwards-incompatible for OSS deployments that previously called these routes with non-admin keys; deployments that want the old behavior can override with HeaderAuthProvider(operation_access={...}). Storage namespace continues to come from get_namespace_key, matching the bindings migration in #204. The unified principal-derived cutover across /controls, /policies, /agents, and /evaluation is a follow-up.
…utes Move auth-framework rationale on /controls/schema, /controls/validate, and /control-templates/render from route docstrings into normal code comments. The docstrings flow into the generated TypeScript SDK as public API documentation, so internal terminology like ``require_operation`` and "upstream authorizer" should not appear there. Function-level comments preserve the rationale for readers of the source. Also remove the skipped placeholder test for the project-scoped credential deny scenario; that scenario depends on a deployment-side provider configuration that is not part of the OSS server, so tracking it as a permanent skipped test in this repo was the wrong home for it. Regenerate the TypeScript SDK to drop the leaked rationale lines.
Add explicit none, api_key, and jwt runtime auth modes, including a generic no-auth provider. Move controls, bindings, policies, agents, and evaluation storage lookups onto principal namespace scoping. Cover auth mode selection and principal namespace isolation with server tests.
…stream The default forward set (X-API-Key, Authorization, Cookie) only covers credential headers Agent Control itself reads. Deployments whose upstream authenticates against a different header name (e.g., a deployer-specific API-key header) had no way to surface that credential through HttpUpstreamAuthProvider — the inbound header reached AC but never crossed the upstream call. Add an extra_forward_headers config field on HttpUpstreamConfig (defaulting to the empty tuple) that operators populate via the new AGENT_CONTROL_AUTH_UPSTREAM_EXTRA_FORWARD_HEADERS env var (comma- separated). The provider's _forward_headers iterates over the union of the default set and the extras, deduplicating case-insensitively so a duplicate name (cross-set or within extras) does not produce two copies on the wire. Tests: - forwards a configured extra header alongside defaults - default forward set unchanged when extras are empty - extras dedupe against defaults case-insensitively - _parse_extra_forward_headers parametric: None / empty / single / multiple / whitespace / empty-entries / case-folded duplicates - configure_auth_from_env threads the parsed tuple onto the provider Lint clean, typecheck clean, full server suite (747) green.
Mirrors #204's bindings migration: replaces require_admin_key and router-level require_api_key with require_operation(CONTROLS_*) on every protected route on /controls and on /control-templates/render. Both routers now mount with the non-validating get_api_key_from_header so the framework owns authentication and authorization, with the extractor attached purely so the generated OpenAPI advertises X-API-Key. GET /controls/schema is intentionally left without a require_operation dependency: it returns a static model schema with no tenant state and routing it through the framework would force the upstream provider to handle a meta-only operation that has no permission semantics. POST /controls/validate and POST /control-templates/render are wired to CONTROLS_CREATE rather than CONTROLS_READ. Both exercise the authoring materialization path and exist to support the create / set- data flow; a caller who cannot create controls has no use for the result. Backwards-incompatible for OSS deployments that previously called these routes with non-admin keys; deployments that want the old behavior can override with HeaderAuthProvider(operation_access={...}). Storage namespace continues to come from get_namespace_key, matching the bindings migration in #204. The unified principal-derived cutover across /controls, /policies, /agents, and /evaluation is a follow-up.
…utes Move auth-framework rationale on /controls/schema, /controls/validate, and /control-templates/render from route docstrings into normal code comments. The docstrings flow into the generated TypeScript SDK as public API documentation, so internal terminology like ``require_operation`` and "upstream authorizer" should not appear there. Function-level comments preserve the rationale for readers of the source. Also remove the skipped placeholder test for the project-scoped credential deny scenario; that scenario depends on a deployment-side provider configuration that is not part of the OSS server, so tracking it as a permanent skipped test in this repo was the wrong home for it. Regenerate the TypeScript SDK to drop the leaked rationale lines.
Add explicit none, api_key, and jwt runtime auth modes, including a generic no-auth provider. Move controls, bindings, policies, agents, and evaluation storage lookups onto principal namespace scoping. Cover auth mode selection and principal namespace isolation with server tests.
…stream The default forward set (X-API-Key, Authorization, Cookie) only covers credential headers Agent Control itself reads. Deployments whose upstream authenticates against a different header name (e.g., a deployer-specific API-key header) had no way to surface that credential through HttpUpstreamAuthProvider — the inbound header reached AC but never crossed the upstream call. Add an extra_forward_headers config field on HttpUpstreamConfig (defaulting to the empty tuple) that operators populate via the new AGENT_CONTROL_AUTH_UPSTREAM_EXTRA_FORWARD_HEADERS env var (comma- separated). The provider's _forward_headers iterates over the union of the default set and the extras, deduplicating case-insensitively so a duplicate name (cross-set or within extras) does not produce two copies on the wire. Tests: - forwards a configured extra header alongside defaults - default forward set unchanged when extras are empty - extras dedupe against defaults case-insensitively - _parse_extra_forward_headers parametric: None / empty / single / multiple / whitespace / empty-entries / case-folded duplicates - configure_auth_from_env threads the parsed tuple onto the provider Lint clean, typecheck clean, full server suite (747) green.
Exchange target-bound runtime tokens for evaluation requests when configured, cache them per target, and retry once after a 401. Keep no-auth and API-key runtime flows on the existing request-auth path when token exchange is unavailable or disabled.
Default stays X-API-Key; pass api_key_header=... or set AGENT_CONTROL_API_KEY_HEADER to override when the upstream auth expects a different header.
Codecov Report❌ Patch coverage is 📢 Thoughts on this report? Let us know! |
namrataghadi-galileo
changed the title
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
test luna and agent control end to end in galileo cluster