feat(pi): add azure subprovider support with base_url config#921
Merged
feat(pi): add azure subprovider support with base_url config#921
Conversation
Add azure-openai-responses as a supported subprovider for pi-coding-agent and pi-cli providers. Users can now configure Azure OpenAI entirely via targets.yaml using `subprovider: azure` (short alias) plus `base_url`, `model`, and `api_key` fields. - Extract shared pi-provider-aliases module (ENV_KEY_MAP, ENV_BASE_URL_MAP, resolveSubprovider) to deduplicate provider-to-env-var mapping - Add baseUrl field to PiCodingAgentResolvedConfig and PiCliResolvedConfig - Resolve base_url/baseUrl/endpoint from target YAML into the config - Set AZURE_OPENAI_BASE_URL env var from config so the pi-ai SDK picks it up Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
… fix - Add `azure-v1` alias for Azure /v1 endpoints that use the standard OpenAI client (resolves to `openai-responses`, avoids AzureOpenAI api-version conflicts) - Construct fallback model descriptor when getModel returns undefined, enabling Azure deployment names not in the pi-ai registry - Set model.baseUrl from config so the SDK uses the correct endpoint - Call registerBuiltInApiProviders() to ensure streaming providers are available - Fix pi-cli spawn on Windows with `shell: process.platform === 'win32'` - Add `azure-v1` to PROVIDER_OWN_PREFIXES for correct env var stripping Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The function exists at runtime in pi-ai but is not in the type definitions. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Deploying agentv with
|
| Latest commit: |
7ea8628
|
| Status: | ⚡️ Build in progress... |
- Resolve npm .cmd wrappers to underlying node scripts on Windows, avoiding shell escaping issues with prompt content containing special characters (<, *, -, etc.) - Add resolveCliProvider() for pi-cli --provider flag mapping (azure-v1 → openai, azure → azure-openai-responses) - Pi CLI uses different provider names than the SDK, so the CLI and SDK alias maps are separate Verified e2e: pi-cli-azure target successfully connects to Azure /v1 endpoint with gpt-5.4-mini, all 3 eval tests execute without errors. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Import accessSync and readFileSync from node:fs - Fix import formatting for biome Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Instead of requiring users to know about azure-v1 vs azure, the
subprovider: azure now automatically uses the OpenAI-compatible client
when base_url is provided. This means azure /v1 endpoints work out of
the box with just:
subprovider: azure
base_url: ${{ AZURE_OPENAI_ENDPOINT }}
When base_url is absent, falls back to AzureOpenAI client with
AZURE_OPENAI_RESOURCE_NAME.
- Replace static alias maps with base_url-aware resolve functions
- Add resolveEnvKeyName/resolveEnvBaseUrlName helpers
- Update PROVIDER_OWN_PREFIXES to use resolved CLI provider name
- Update targets.yaml to use AZURE_OPENAI_ENDPOINT directly
- Comprehensive tests for all resolve functions with/without base_url
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Pi-cli azure now uses the native azure-openai-responses provider with AZURE_OPENAI_RESOURCE_NAME (extracted from base_url). This means subprovider: azure works out of the box — no azure-v1 alias needed. - extractAzureResourceName() parses resource name from URLs or passes through raw names - Pi-cli buildEnv sets AZURE_OPENAI_API_KEY + AZURE_OPENAI_RESOURCE_NAME - resolveCliProvider always maps azure → azure-openai-responses - SDK path unchanged: azure + base_url → openai-responses (v1 compat) - Add azure-smoke.eval.yaml for e2e testing Verified e2e: pi CLI connects to Azure, model returns correct answers (Paris, 2+2=4). Response parsing is a separate pre-existing issue. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…r azure - resolveWindowsCmd now uses `where` to find the full path of the .cmd wrapper, fixing resolution when the executable isn't in CWD - Skip --api-key flag for azure subprovider since the key is passed via AZURE_OPENAI_API_KEY env var (--api-key sets the wrong provider's key) Note: pi-cli stdout capture on Windows is broken due to a Bun runtime bug (Bun.spawn/child_process.spawn can't capture stdout from node child processes on Windows). This affects all pi-cli evals on Windows, not just azure. The azure connectivity itself is verified working — the pi CLI returns correct responses (Paris, 4) but Bun can't read them. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Some providers (azure-openai-responses) may emit text content in message_update events but leave the agent_end assistant message with empty content. Fall back to the last message_end event with non-empty content. Also adds azure-smoke.eval.yaml for e2e connectivity testing. Note: pi-cli azure evals return empty assistant content due to the pi-coding-agent CLI's azure-openai-responses streaming not populating the agent_end messages array. This is an upstream pi-ai issue. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
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
1 participant
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.
Summary
azure-openai-responsesas a supported subprovider forpi-coding-agentandpi-cliproviders via the short aliassubprovider: azurebase_urlfield to both pi provider configs, enabling full Azure OpenAI configuration viatargets.yamlwithout requiring ambient env varspi-provider-aliasesmodule to deduplicate provider-to-env-var mapping (ENV_KEY_MAP,ENV_BASE_URL_MAP,resolveSubprovider)Example target config
Test plan
resolveSubprovideralias resolution (azure → azure-openai-responses)ENV_KEY_MAPandENV_BASE_URL_MAPmappingstargets.test.tsAGENT_TARGET=pi-azureagainst Azure OpenAI endpoint🤖 Generated with Claude Code