Merge branch 'main' into node24

.github/workflows/release-new-action-version.yml

@@ -16,6 +16,6 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Update the ${{ env.TAG_NAME }} tag
-        uses: actions/publish-action@v0.2.2
+        uses: actions/publish-action@v0.3.0
         with:
           source-tag: ${{ env.TAG_NAME }}

.licenses/npm/openai.dep.yml

@@ -0,0 +1,212 @@
+---
+name: openai
+version: 5.11.0
+type: npm
+summary: The official TypeScript library for the OpenAI API
+homepage:
+license: apache-2.0
+licenses:
+- sources: LICENSE
+  text: |2
+                                     Apache License
+                               Version 2.0, January 2004
+                            http://www.apache.org/licenses/
+
+       TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+       1. Definitions.
+
+          "License" shall mean the terms and conditions for use, reproduction,
+          and distribution as defined by Sections 1 through 9 of this document.
+
+          "Licensor" shall mean the copyright owner or entity authorized by
+          the copyright owner that is granting the License.
+
+          "Legal Entity" shall mean the union of the acting entity and all
+          other entities that control, are controlled by, or are under common
+          control with that entity. For the purposes of this definition,
+          "control" means (i) the power, direct or indirect, to cause the
+          direction or management of such entity, whether by contract or
+          otherwise, or (ii) ownership of fifty percent (50%) or more of the
+          outstanding shares, or (iii) beneficial ownership of such entity.
+
+          "You" (or "Your") shall mean an individual or Legal Entity
+          exercising permissions granted by this License.
+
+          "Source" form shall mean the preferred form for making modifications,
+          including but not limited to software source code, documentation
+          source, and configuration files.
+
+          "Object" form shall mean any form resulting from mechanical
+          transformation or translation of a Source form, including but
+          not limited to compiled object code, generated documentation,
+          and conversions to other media types.
+
+          "Work" shall mean the work of authorship, whether in Source or
+          Object form, made available under the License, as indicated by a
+          copyright notice that is included in or attached to the work
+          (an example is provided in the Appendix below).
+
+          "Derivative Works" shall mean any work, whether in Source or Object
+          form, that is based on (or derived from) the Work and for which the
+          editorial revisions, annotations, elaborations, or other modifications
+          represent, as a whole, an original work of authorship. For the purposes
+          of this License, Derivative Works shall not include works that remain
+          separable from, or merely link (or bind by name) to the interfaces of,
+          the Work and Derivative Works thereof.
+
+          "Contribution" shall mean any work of authorship, including
+          the original version of the Work and any modifications or additions
+          to that Work or Derivative Works thereof, that is intentionally
+          submitted to Licensor for inclusion in the Work by the copyright owner
+          or by an individual or Legal Entity authorized to submit on behalf of
+          the copyright owner. For the purposes of this definition, "submitted"
+          means any form of electronic, verbal, or written communication sent
+          to the Licensor or its representatives, including but not limited to
+          communication on electronic mailing lists, source code control systems,
+          and issue tracking systems that are managed by, or on behalf of, the
+          Licensor for the purpose of discussing and improving the Work, but
+          excluding communication that is conspicuously marked or otherwise
+          designated in writing by the copyright owner as "Not a Contribution."
+
+          "Contributor" shall mean Licensor and any individual or Legal Entity
+          on behalf of whom a Contribution has been received by Licensor and
+          subsequently incorporated within the Work.
+
+       2. Grant of Copyright License. Subject to the terms and conditions of
+          this License, each Contributor hereby grants to You a perpetual,
+          worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+          copyright license to reproduce, prepare Derivative Works of,
+          publicly display, publicly perform, sublicense, and distribute the
+          Work and such Derivative Works in Source or Object form.
+
+       3. Grant of Patent License. Subject to the terms and conditions of
+          this License, each Contributor hereby grants to You a perpetual,
+          worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+          (except as stated in this section) patent license to make, have made,
+          use, offer to sell, sell, import, and otherwise transfer the Work,
+          where such license applies only to those patent claims licensable
+          by such Contributor that are necessarily infringed by their
+          Contribution(s) alone or by combination of their Contribution(s)
+          with the Work to which such Contribution(s) was submitted. If You
+          institute patent litigation against any entity (including a
+          cross-claim or counterclaim in a lawsuit) alleging that the Work
+          or a Contribution incorporated within the Work constitutes direct
+          or contributory patent infringement, then any patent licenses
+          granted to You under this License for that Work shall terminate
+          as of the date such litigation is filed.
+
+       4. Redistribution. You may reproduce and distribute copies of the
+          Work or Derivative Works thereof in any medium, with or without
+          modifications, and in Source or Object form, provided that You
+          meet the following conditions:
+
+          (a) You must give any other recipients of the Work or
+              Derivative Works a copy of this License; and
+
+          (b) You must cause any modified files to carry prominent notices
+              stating that You changed the files; and
+
+          (c) You must retain, in the Source form of any Derivative Works
+              that You distribute, all copyright, patent, trademark, and
+              attribution notices from the Source form of the Work,
+              excluding those notices that do not pertain to any part of
+              the Derivative Works; and
+
+          (d) If the Work includes a "NOTICE" text file as part of its
+              distribution, then any Derivative Works that You distribute must
+              include a readable copy of the attribution notices contained
+              within such NOTICE file, excluding those notices that do not
+              pertain to any part of the Derivative Works, in at least one
+              of the following places: within a NOTICE text file distributed
+              as part of the Derivative Works; within the Source form or
+              documentation, if provided along with the Derivative Works; or,
+              within a display generated by the Derivative Works, if and
+              wherever such third-party notices normally appear. The contents
+              of the NOTICE file are for informational purposes only and
+              do not modify the License. You may add Your own attribution
+              notices within Derivative Works that You distribute, alongside
+              or as an addendum to the NOTICE text from the Work, provided
+              that such additional attribution notices cannot be construed
+              as modifying the License.
+
+          You may add Your own copyright statement to Your modifications and
+          may provide additional or different license terms and conditions
+          for use, reproduction, or distribution of Your modifications, or
+          for any such Derivative Works as a whole, provided Your use,
+          reproduction, and distribution of the Work otherwise complies with
+          the conditions stated in this License.
+
+       5. Submission of Contributions. Unless You explicitly state otherwise,
+          any Contribution intentionally submitted for inclusion in the Work
+          by You to the Licensor shall be under the terms and conditions of
+          this License, without any additional terms or conditions.
+          Notwithstanding the above, nothing herein shall supersede or modify
+          the terms of any separate license agreement you may have executed
+          with Licensor regarding such Contributions.
+
+       6. Trademarks. This License does not grant permission to use the trade
+          names, trademarks, service marks, or product names of the Licensor,
+          except as required for reasonable and customary use in describing the
+          origin of the Work and reproducing the content of the NOTICE file.
+
+       7. Disclaimer of Warranty. Unless required by applicable law or
+          agreed to in writing, Licensor provides the Work (and each
+          Contributor provides its Contributions) on an "AS IS" BASIS,
+          WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+          implied, including, without limitation, any warranties or conditions
+          of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+          PARTICULAR PURPOSE. You are solely responsible for determining the
+          appropriateness of using or redistributing the Work and assume any
+          risks associated with Your exercise of permissions under this License.
+
+       8. Limitation of Liability. In no event and under no legal theory,
+          whether in tort (including negligence), contract, or otherwise,
+          unless required by applicable law (such as deliberate and grossly
+          negligent acts) or agreed to in writing, shall any Contributor be
+          liable to You for damages, including any direct, indirect, special,
+          incidental, or consequential damages of any character arising as a
+          result of this License or out of the use or inability to use the
+          Work (including but not limited to damages for loss of goodwill,
+          work stoppage, computer failure or malfunction, or any and all
+          other commercial damages or losses), even if such Contributor
+          has been advised of the possibility of such damages.
+
+       9. Accepting Warranty or Additional Liability. While redistributing
+          the Work or Derivative Works thereof, You may choose to offer,
+          and charge a fee for, acceptance of support, warranty, indemnity,
+          or other liability obligations and/or rights consistent with this
+          License. However, in accepting such obligations, You may act only
+          on Your own behalf and on Your sole responsibility, not on behalf
+          of any other Contributor, and only if You agree to indemnify,
+          defend, and hold each Contributor harmless for any liability
+          incurred by, or claims asserted against, such Contributor by reason
+          of your accepting any such warranty or additional liability.
+
+       END OF TERMS AND CONDITIONS
+
+       APPENDIX: How to apply the Apache License to your work.
+
+          To apply the Apache License to your work, attach the following
+          boilerplate notice, with the fields enclosed by brackets "[]"
+          replaced with your own identifying information. (Don't include
+          the brackets!)  The text should be enclosed in the appropriate
+          comment syntax for the file format. We also recommend that a
+          file or class name and description of purpose be included on the
+          same "printed page" as the copyright notice for easier
+          identification within third-party archives.
+
+       Copyright 2025 OpenAI
+
+       Licensed under the Apache License, Version 2.0 (the "License");
+       you may not use this file except in compliance with the License.
+       You may obtain a copy of the License at
+
+           http://www.apache.org/licenses/LICENSE-2.0
+
+       Unless required by applicable law or agreed to in writing, software
+       distributed under the License is distributed on an "AS IS" BASIS,
+       WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+       See the License for the specific language governing permissions and
+       limitations under the License.
+notices: []

README.md

@@ -65,6 +65,9 @@ steps:
         var3: |
           Lorem Ipsum
           Hello World
+      file_input: |
+        var4: ./path/to/long-text.txt
+        var5: ./path/to/config.json
 ```
 
 #### Simple prompt.yml example
@@ -116,7 +119,9 @@ jsonSchema: |-
 ```
 
 Variables in prompt.yml files are templated using `{{variable}}` format and are
-supplied via the `input` parameter in YAML format.
+supplied via the `input` parameter in YAML format. Additionally, you can
+provide file-based variables via `file_input`, where each key maps to a file
+path.
 
 ### Using a system prompt file
 
@@ -168,12 +173,24 @@ steps:
       token: ${{ secrets.USER_PAT }}
 ```
 
+If you want, you can use separate tokens for the AI inference endpoint
+and the GitHub MCP server:
+
+```yaml
+steps:
+  - name: AI Inference with Separate MCP Token
+    id: inference
+    uses: actions/[email protected]
+    with:
+      prompt: 'List my open pull requests and create a summary'
+      enable-github-mcp: true
+      token: ${{ secrets.GITHUB_TOKEN }}
+      github-mcp-token: ${{ secrets.USER_PAT }}
+```
+
 When MCP is enabled, the AI model will have access to GitHub tools and can
 perform actions like searching issues and PRs.
 
-**Note:** For now, MCP integration cannot be used with the built-in token. You
-must pass a GitHub PAT into `token:` instead.
-
 ## Inputs
 
 Various inputs are defined in [`action.yml`](action.yml) to let you configure
@@ -185,12 +202,14 @@ the action:
 | `prompt`             | The prompt to send to the model                                                                                                                               | N/A                                  |
 | `prompt-file`        | Path to a file containing the prompt (supports .txt and .prompt.yml formats). If both `prompt` and `prompt-file` are provided, `prompt-file` takes precedence | `""`                                 |
 | `input`              | Template variables in YAML format for .prompt.yml files (e.g., `var1: value1` on separate lines)                                                              | `""`                                 |
+| `file_input`         | Template variables in YAML where values are file paths. The file contents are read and used for templating                                                    | `""`                                 |
 | `system-prompt`      | The system prompt to send to the model                                                                                                                        | `"You are a helpful assistant"`      |
 | `system-prompt-file` | Path to a file containing the system prompt. If both `system-prompt` and `system-prompt-file` are provided, `system-prompt-file` takes precedence             | `""`                                 |
 | `model`              | The model to use for inference. Must be available in the [GitHub Models](https://github.com/marketplace?type=models) catalog                                  | `openai/gpt-4o`                      |
 | `endpoint`           | The endpoint to use for inference. If you're running this as part of an org, you should probably use the org-specific Models endpoint                         | `https://models.github.ai/inference` |
 | `max-tokens`         | The max number of tokens to generate                                                                                                                          | 200                                  |
 | `enable-github-mcp`  | Enable Model Context Protocol integration with GitHub tools                                                                                                   | `false`                              |
+| `github-mcp-token`   | Token to use for GitHub MCP server (defaults to the main token if not specified). Use a separate PAT for tighter security                                     | `""`                                 |
 
 ## Outputs
 

__tests__/inference.test.ts

@@ -2,17 +2,15 @@ import {vi, type MockedFunction, beforeEach, expect, describe, it} from 'vitest'
 import * as core from '../__fixtures__/core.js'
 
 // eslint-disable-next-line @typescript-eslint/no-explicit-any
-const mockPost = vi.fn() as MockedFunction<any>
+const mockCreate = vi.fn() as MockedFunction<any>
-const mockPath = vi.fn(() => ({post: mockPost}))
+const mockCompletions = {create: mockCreate}
-const mockClient = vi.fn(() => ({path: mockPath}))
+const mockChat = {completions: mockCompletions}
-
+const mockOpenAIClient = vi.fn(() => ({
-vi.mock('@azure-rest/ai-inference', () => ({
+  chat: mockChat,
-  default: mockClient,
-  isUnexpected: vi.fn(() => false),
 }))
 
-vi.mock('@azure/core-auth', () => ({
+vi.mock('openai', () => ({
-  AzureKeyCredential: vi.fn(),
+  default: mockOpenAIClient,
 }))
 
 // eslint-disable-next-line @typescript-eslint/no-explicit-any
@@ -29,8 +27,8 @@ const {simpleInference, mcpInference} = await import('../src/inference.js')
 describe('inference.ts', () => {
   const mockRequest = {
     messages: [
-      {role: 'system', content: 'You are a test assistant'},
+      {role: 'system' as const, content: 'You are a test assistant'},
-      {role: 'user', content: 'Hello, AI!'},
+      {role: 'user' as const, content: 'Hello, AI!'},
     ],
     modelName: 'gpt-4',
     maxTokens: 100,
@@ -45,18 +43,16 @@ describe('inference.ts', () => {
   describe('simpleInference', () => {
     it('performs simple inference without tools', async () => {
       const mockResponse = {
-        body: {
+        choices: [
-          choices: [
+          {
-            {
+            message: {
-              message: {
+              content: 'Hello, user!',
-                content: 'Hello, user!',
-              },
             },
-          ],
+          },
-        },
+        ],
       }
 
-      mockPost.mockResolvedValue(mockResponse)
+      mockCreate.mockResolvedValue(mockResponse)
 
       const result = await simpleInference(mockRequest)
 
@@ -65,44 +61,83 @@ describe('inference.ts', () => {
       expect(core.info).toHaveBeenCalledWith('Model response: Hello, user!')
 
       // Verify the request structure
-      expect(mockPost).toHaveBeenCalledWith({
+      expect(mockCreate).toHaveBeenCalledWith({
-        body: {
+        messages: [
-          messages: [
+          {
-            {
+            role: 'system',
-              role: 'system',
+            content: 'You are a test assistant',
-              content: 'You are a test assistant',
+          },
-            },
+          {
-            {
+            role: 'user',
-              role: 'user',
+            content: 'Hello, AI!',
-              content: 'Hello, AI!',
+          },
-            },
+        ],
-          ],
+        max_tokens: 100,
-          max_tokens: 100,
+        model: 'gpt-4',
-          model: 'gpt-4',
-        },
       })
     })
 
     it('handles null response content', async () => {
       const mockResponse = {
-        body: {
+        choices: [
-          choices: [
+          {
-            {
+            message: {
-              message: {
+              content: null,
-                content: null,
-              },
             },
-          ],
+          },
-        },
+        ],
       }
 
-      mockPost.mockResolvedValue(mockResponse)
+      mockCreate.mockResolvedValue(mockResponse)
 
       const result = await simpleInference(mockRequest)
 
       expect(result).toBeNull()
       expect(core.info).toHaveBeenCalledWith('Model response: No response content')
     })
+
+    it('includes response format when specified', async () => {
+      const requestWithResponseFormat = {
+        ...mockRequest,
+        responseFormat: {
+          type: 'json_schema' as const,
+          json_schema: {type: 'object'},
+        },
+      }
+
+      const mockResponse = {
+        choices: [
+          {
+            message: {
+              content: '{"result": "success"}',
+            },
+          },
+        ],
+      }
+
+      mockCreate.mockResolvedValue(mockResponse)
+
+      const result = await simpleInference(requestWithResponseFormat)
+
+      expect(result).toBe('{"result": "success"}')
+
+      // Verify response format was included in the request
+      expect(mockCreate).toHaveBeenCalledWith({
+        messages: [
+          {
+            role: 'system',
+            content: 'You are a test assistant',
+          },
+          {
+            role: 'user',
+            content: 'Hello, AI!',
+          },
+        ],
+        max_tokens: 100,
+        model: 'gpt-4',
+        response_format: requestWithResponseFormat.responseFormat,
+      })
+    })
   })
 
   describe('mcpInference', () => {
@@ -123,19 +158,17 @@ describe('inference.ts', () => {
 
     it('performs inference without tool calls', async () => {
       const mockResponse = {
-        body: {
+        choices: [
-          choices: [
+          {
-            {
+            message: {
-              message: {
+              content: 'Hello, user!',
-                content: 'Hello, user!',
+              tool_calls: null,
-                tool_calls: null,
-              },
             },
-          ],
+          },
-        },
+        ],
       }
 
-      mockPost.mockResolvedValue(mockResponse)
+      mockCreate.mockResolvedValue(mockResponse)
 
       const result = await mcpInference(mockRequest, mockMcpClient)
 
@@ -146,12 +179,13 @@ describe('inference.ts', () => {
 
       // The MCP inference loop will always add the assistant message, even when there are no tool calls
       // So we don't check the exact messages, just that tools were included
-      expect(mockPost).toHaveBeenCalledTimes(1)
+      expect(mockCreate).toHaveBeenCalledTimes(1)
       // eslint-disable-next-line @typescript-eslint/no-explicit-any
-      const callArgs = mockPost.mock.calls[0][0] as any
+      const callArgs = mockCreate.mock.calls[0][0] as any
-      expect(callArgs.body.tools).toEqual(mockMcpClient.tools)
+      expect(callArgs.tools).toEqual(mockMcpClient.tools)
-      expect(callArgs.body.model).toBe('gpt-4')
+      expect(callArgs.response_format).toBeUndefined()
-      expect(callArgs.body.max_tokens).toBe(100)
+      expect(callArgs.model).toBe('gpt-4')
+      expect(callArgs.max_tokens).toBe(100)
     })
 
     it('executes tool calls and continues conversation', async () => {
@@ -176,33 +210,29 @@ describe('inference.ts', () => {
 
       // First response with tool calls
       const firstResponse = {
-        body: {
+        choices: [
-          choices: [
+          {
-            {
+            message: {
-              message: {
+              content: 'I need to use a tool.',
-                content: 'I need to use a tool.',
+              tool_calls: toolCalls,
-                tool_calls: toolCalls,
-              },
             },
-          ],
+          },
-        },
+        ],
       }
 
       // Second response after tool execution
       const secondResponse = {
-        body: {
+        choices: [
-          choices: [
+          {
-            {
+            message: {
-              message: {
+              content: 'Here is the final answer.',
-                content: 'Here is the final answer.',
+              tool_calls: null,
-                tool_calls: null,
-              },
             },
-          ],
+          },
-        },
+        ],
       }
 
-      mockPost.mockResolvedValueOnce(firstResponse).mockResolvedValueOnce(secondResponse)
+      mockCreate.mockResolvedValueOnce(firstResponse).mockResolvedValueOnce(secondResponse)
 
       mockExecuteToolCalls.mockResolvedValue(toolResults)
 
@@ -210,15 +240,15 @@ describe('inference.ts', () => {
 
       expect(result).toBe('Here is the final answer.')
       expect(mockExecuteToolCalls).toHaveBeenCalledWith(mockMcpClient.client, toolCalls)
-      expect(mockPost).toHaveBeenCalledTimes(2)
+      expect(mockCreate).toHaveBeenCalledTimes(2)
 
       // Verify the second call includes the conversation history
       // eslint-disable-next-line @typescript-eslint/no-explicit-any
-      const secondCall = mockPost.mock.calls[1][0] as any
+      const secondCall = mockCreate.mock.calls[1][0] as any
-      expect(secondCall.body.messages).toHaveLength(5) // system, user, assistant, tool, assistant
+      expect(secondCall.messages).toHaveLength(5) // system, user, assistant, tool, assistant
-      expect(secondCall.body.messages[2].role).toBe('assistant')
+      expect(secondCall.messages[2].role).toBe('assistant')
-      expect(secondCall.body.messages[2].tool_calls).toEqual(toolCalls)
+      expect(secondCall.messages[2].tool_calls).toEqual(toolCalls)
-      expect(secondCall.body.messages[3]).toEqual(toolResults[0])
+      expect(secondCall.messages[3]).toEqual(toolResults[0])
     })
 
     it('handles maximum iteration limit', async () => {
@@ -243,43 +273,39 @@ describe('inference.ts', () => {
 
       // Always respond with tool calls to trigger infinite loop
       const responseWithToolCalls = {
-        body: {
+        choices: [
-          choices: [
+          {
-            {
+            message: {
-              message: {
+              content: 'Using tool again.',
-                content: 'Using tool again.',
+              tool_calls: toolCalls,
-                tool_calls: toolCalls,
-              },
             },
-          ],
+          },
-        },
+        ],
       }
 
-      mockPost.mockResolvedValue(responseWithToolCalls)
+      mockCreate.mockResolvedValue(responseWithToolCalls)
       mockExecuteToolCalls.mockResolvedValue(toolResults)
 
       const result = await mcpInference(mockRequest, mockMcpClient)
 
-      expect(mockPost).toHaveBeenCalledTimes(5) // Max iterations reached
+      expect(mockCreate).toHaveBeenCalledTimes(5) // Max iterations reached
       expect(core.warning).toHaveBeenCalledWith('GitHub MCP inference loop exceeded maximum iterations (5)')
       expect(result).toBe('Using tool again.') // Last assistant message
     })
 
     it('handles empty tool calls array', async () => {
       const mockResponse = {
-        body: {
+        choices: [
-          choices: [
+          {
-            {
+            message: {
-              message: {
+              content: 'Hello, user!',
-                content: 'Hello, user!',
+              tool_calls: [],
-                tool_calls: [],
-              },
             },
-          ],
+          },
-        },
+        ],
       }
 
-      mockPost.mockResolvedValue(mockResponse)
+      mockCreate.mockResolvedValue(mockResponse)
 
       const result = await mcpInference(mockRequest, mockMcpClient)
 
@@ -297,32 +323,28 @@ describe('inference.ts', () => {
       ]
 
       const firstResponse = {
-        body: {
+        choices: [
-          choices: [
+          {
-            {
+            message: {
-              message: {
+              content: 'First message',
-                content: 'First message',
+              tool_calls: toolCalls,
-                tool_calls: toolCalls,
-              },
             },
-          ],
+          },
-        },
+        ],
       }
 
       const secondResponse = {
-        body: {
+        choices: [
-          choices: [
+          {
-            {
+            message: {
-              message: {
+              content: 'Second message',
-                content: 'Second message',
+              tool_calls: toolCalls,
-                tool_calls: toolCalls,
-              },
             },
-          ],
+          },
-        },
+        ],
       }
 
-      mockPost.mockResolvedValueOnce(firstResponse).mockResolvedValue(secondResponse)
+      mockCreate.mockResolvedValueOnce(firstResponse).mockResolvedValue(secondResponse)
 
       mockExecuteToolCalls.mockResolvedValue([
         {
@@ -337,5 +359,191 @@ describe('inference.ts', () => {
 
       expect(result).toBe('Second message')
     })
+
+    it('makes additional loop with response format when no tool calls are made', async () => {
+      const requestWithResponseFormat = {
+        ...mockRequest,
+        responseFormat: {
+          type: 'json_schema' as const,
+          json_schema: {type: 'object'},
+        },
+      }
+
+      // First response without tool calls
+      const firstResponse = {
+        choices: [
+          {
+            message: {
+              content: 'First response',
+              tool_calls: null,
+            },
+          },
+        ],
+      }
+
+      // Second response with response format applied
+      const secondResponse = {
+        choices: [
+          {
+            message: {
+              content: '{"result": "formatted response"}',
+              tool_calls: null,
+            },
+          },
+        ],
+      }
+
+      mockCreate.mockResolvedValueOnce(firstResponse).mockResolvedValueOnce(secondResponse)
+
+      const result = await mcpInference(requestWithResponseFormat, mockMcpClient)
+
+      expect(result).toBe('{"result": "formatted response"}')
+      expect(mockCreate).toHaveBeenCalledTimes(2)
+      expect(core.info).toHaveBeenCalledWith('Making one more MCP loop with the requested response format...')
+
+      // First call should have tools but no response format
+      // eslint-disable-next-line @typescript-eslint/no-explicit-any
+      const firstCall = mockCreate.mock.calls[0][0] as any
+      expect(firstCall.tools).toEqual(mockMcpClient.tools)
+      expect(firstCall.response_format).toBeUndefined()
+
+      // Second call should have response format but no tools
+      // eslint-disable-next-line @typescript-eslint/no-explicit-any
+      const secondCall = mockCreate.mock.calls[1][0] as any
+      expect(secondCall.tools).toBeUndefined()
+      expect(secondCall.response_format).toEqual(requestWithResponseFormat.responseFormat)
+
+      // Second call should include the user message requesting JSON format
+      expect(secondCall.messages).toHaveLength(5) // system, user, assistant, user, assistant
+      expect(secondCall.messages[3].role).toBe('user')
+      expect(secondCall.messages[3].content).toContain('Please provide your response in the exact')
+    })
+
+    it('uses response format only on final iteration after tool calls', async () => {
+      const requestWithResponseFormat = {
+        ...mockRequest,
+        responseFormat: {
+          type: 'json_schema' as const,
+          json_schema: {type: 'object'},
+        },
+      }
+
+      const toolCalls = [
+        {
+          id: 'call-123',
+          function: {
+            name: 'test-tool',
+            arguments: '{"param": "value"}',
+          },
+        },
+      ]
+
+      const toolResults = [
+        {
+          tool_call_id: 'call-123',
+          role: 'tool',
+          name: 'test-tool',
+          content: 'Tool result',
+        },
+      ]
+
+      // First response with tool calls
+      const firstResponse = {
+        choices: [
+          {
+            message: {
+              content: 'Using tool',
+              tool_calls: toolCalls,
+            },
+          },
+        ],
+      }
+
+      // Second response without tool calls, but should trigger final message loop
+      const secondResponse = {
+        choices: [
+          {
+            message: {
+              content: 'Intermediate result',
+              tool_calls: null,
+            },
+          },
+        ],
+      }
+
+      // Third response with response format
+      const thirdResponse = {
+        choices: [
+          {
+            message: {
+              content: '{"final": "result"}',
+              tool_calls: null,
+            },
+          },
+        ],
+      }
+
+      mockCreate
+        .mockResolvedValueOnce(firstResponse)
+        .mockResolvedValueOnce(secondResponse)
+        .mockResolvedValueOnce(thirdResponse)
+
+      mockExecuteToolCalls.mockResolvedValue(toolResults)
+
+      const result = await mcpInference(requestWithResponseFormat, mockMcpClient)
+
+      expect(result).toBe('{"final": "result"}')
+      expect(mockCreate).toHaveBeenCalledTimes(3)
+
+      // First call: tools but no response format
+      // eslint-disable-next-line @typescript-eslint/no-explicit-any
+      const firstCall = mockCreate.mock.calls[0][0] as any
+      expect(firstCall.tools).toEqual(mockMcpClient.tools)
+      expect(firstCall.response_format).toBeUndefined()
+
+      // Second call: tools but no response format (after tool execution)
+      // eslint-disable-next-line @typescript-eslint/no-explicit-any
+      const secondCall = mockCreate.mock.calls[1][0] as any
+      expect(secondCall.tools).toEqual(mockMcpClient.tools)
+      expect(secondCall.response_format).toBeUndefined()
+
+      // Third call: response format but no tools (final message)
+      // eslint-disable-next-line @typescript-eslint/no-explicit-any
+      const thirdCall = mockCreate.mock.calls[2][0] as any
+      expect(thirdCall.tools).toBeUndefined()
+      expect(thirdCall.response_format).toEqual(requestWithResponseFormat.responseFormat)
+    })
+
+    it('returns immediately when response format is set and finalMessage is already true', async () => {
+      const requestWithResponseFormat = {
+        ...mockRequest,
+        responseFormat: {
+          type: 'json_schema' as const,
+          json_schema: {type: 'object'},
+        },
+      }
+
+      // Response without tool calls on what would be the final message iteration
+      const mockResponse = {
+        choices: [
+          {
+            message: {
+              content: '{"immediate": "result"}',
+              tool_calls: null,
+            },
+          },
+        ],
+      }
+
+      mockCreate.mockResolvedValue(mockResponse)
+
+      // We need to test a scenario where finalMessage would already be true
+      // This happens when we're already in the final iteration
+      const result = await mcpInference(requestWithResponseFormat, mockMcpClient)
+
+      // The function should make two calls: one normal, then one with response format
+      expect(mockCreate).toHaveBeenCalledTimes(2)
+      expect(result).toBe('{"immediate": "result"}')
+    })
   })
 })

__tests__/main-prompt-integration.test.ts

@@ -34,6 +34,12 @@ vi.mock('../src/mcp.js', () => ({
 
 vi.mock('@actions/core', () => core)
 
+// Mock process.exit to prevent it from actually exiting during tests
+const mockProcessExit = vi.spyOn(process, 'exit').mockImplementation(() => {
+  // Prevent actual exit, but don't throw - just return
+  return undefined as never
+})
+
 // The module being tested should be imported dynamically. This ensures that the
 // mocks are used in place of any actual dependencies.
 const {run} = await import('../src/main.js')
@@ -41,6 +47,7 @@ const {run} = await import('../src/main.js')
 describe('main.ts - prompt.yml integration', () => {
   beforeEach(() => {
     vi.clearAllMocks()
+    mockProcessExit.mockClear()
 
     // Mock environment variables
     process.env['GITHUB_TOKEN'] = 'test-token'
@@ -103,8 +110,12 @@ model: openai/gpt-4o
       }
     })
 
+    // Expect the run function to complete successfully
     await run()
 
+    // Verify process.exit was called with code 0 (success)
+    expect(mockProcessExit).toHaveBeenCalledWith(0)
+
     // Verify simpleInference was called with the correct message structure
     expect(mockSimpleInference).toHaveBeenCalledWith(
       expect.objectContaining({
@@ -130,6 +141,52 @@ model: openai/gpt-4o
     expect(core.setOutput).toHaveBeenCalledWith('response-file', expect.any(String))
   })
 
+  it('supports file_input variables to load file contents', async () => {
+    mockExistsSync.mockReturnValue(true)
+
+    // First call: reading the prompt file. Second call: reading file_input referenced file contents.
+    const externalFilePath = 'vars.txt'
+    mockReadFileSync.mockImplementation((path: string) => {
+      if (path === 'test.prompt.yml') {
+        return `messages:\n  - role: user\n    content: 'Here is the data: {{blob}}'\nmodel: openai/gpt-4o\n`
+      }
+      if (path === externalFilePath) {
+        return 'FILE_CONTENTS'
+      }
+      return ''
+    })
+
+    core.getInput.mockImplementation((name: string) => {
+      switch (name) {
+        case 'prompt-file':
+          return 'test.prompt.yml'
+        case 'file_input':
+          return `blob: ${externalFilePath}`
+        case 'model':
+          return 'openai/gpt-4o'
+        case 'max-tokens':
+          return '200'
+        case 'endpoint':
+          return 'https://models.github.ai/inference'
+        case 'enable-github-mcp':
+          return 'false'
+        default:
+          return ''
+      }
+    })
+
+    await run()
+
+    expect(mockSimpleInference).toHaveBeenCalledWith(
+      expect.objectContaining({
+        messages: [{role: 'user', content: 'Here is the data: FILE_CONTENTS'}],
+      }),
+    )
+
+    // Verify process.exit was called with code 0 (success)
+    expect(mockProcessExit).toHaveBeenCalledWith(0)
+  })
+
   it('should fall back to legacy format when not using prompt YAML', async () => {
     mockExistsSync.mockReturnValue(false)
     core.getInput.mockImplementation((name: string) => {
@@ -172,5 +229,8 @@ model: openai/gpt-4o
         token: 'test-token',
       }),
     )
+
+    // Verify process.exit was called with code 0 (success)
+    expect(mockProcessExit).toHaveBeenCalledWith(0)
   })
 })

__tests__/main.test.ts

@@ -94,6 +94,12 @@ vi.mock('../src/inference.js', () => ({
 
 vi.mock('@actions/core', () => core)
 
+// Mock process.exit to prevent it from actually exiting during tests
+const mockProcessExit = vi.spyOn(process, 'exit').mockImplementation(() => {
+  // Prevent actual exit, but don't throw - just return
+  return undefined as never
+})
+
 // The module being tested should be imported dynamically. This ensures that the
 // mocks are used in place of any actual dependencies.
 const {run} = await import('../src/main.js')
@@ -102,6 +108,7 @@ describe('main.ts', () => {
   // Reset all mocks before each test
   beforeEach(() => {
     vi.clearAllMocks()
+    mockProcessExit.mockClear()
 
     // Remove any existing GITHUB_TOKEN
     delete process.env.GITHUB_TOKEN
@@ -121,6 +128,7 @@ describe('main.ts', () => {
 
     expect(core.setOutput).toHaveBeenCalled()
     verifyStandardResponse()
+    expect(mockProcessExit).toHaveBeenCalledWith(0)
   })
 
   it('Sets a failed status when no prompt is set', async () => {
@@ -131,7 +139,8 @@ describe('main.ts', () => {
 
     await run()
 
-    expect(core.setFailed).toHaveBeenNthCalledWith(1, 'Neither prompt-file nor prompt was set')
+    expect(core.setFailed).toHaveBeenCalledWith('Neither prompt-file nor prompt was set')
+    expect(mockProcessExit).toHaveBeenCalledWith(1)
   })
 
   it('uses simple inference when MCP is disabled', async () => {
@@ -157,6 +166,7 @@ describe('main.ts', () => {
     expect(mockConnectToGitHubMCP).not.toHaveBeenCalled()
     expect(mockMcpInference).not.toHaveBeenCalled()
     verifyStandardResponse()
+    expect(mockProcessExit).toHaveBeenCalledWith(0)
   })
 
   it('uses MCP inference when enabled and connection succeeds', async () => {
@@ -189,6 +199,7 @@ describe('main.ts', () => {
     )
     expect(mockSimpleInference).not.toHaveBeenCalled()
     verifyStandardResponse()
+    expect(mockProcessExit).toHaveBeenCalledWith(0)
   })
 
   it('falls back to simple inference when MCP connection fails', async () => {
@@ -207,6 +218,7 @@ describe('main.ts', () => {
     expect(mockMcpInference).not.toHaveBeenCalled()
     expect(core.warning).toHaveBeenCalledWith('MCP connection failed, falling back to simple inference')
     verifyStandardResponse()
+    expect(mockProcessExit).toHaveBeenCalledWith(0)
   })
 
   it('properly integrates with loadContentFromFileOrInput', async () => {
@@ -240,6 +252,7 @@ describe('main.ts', () => {
       responseFormat: undefined,
     })
     verifyStandardResponse()
+    expect(mockProcessExit).toHaveBeenCalledWith(0)
   })
 
   it('handles non-existent prompt-file with an error', async () => {
@@ -254,5 +267,6 @@ describe('main.ts', () => {
     await run()
 
     expect(core.setFailed).toHaveBeenCalledWith(`File for prompt-file was not found: ${promptFile}`)
+    expect(mockProcessExit).toHaveBeenCalledWith(1)
   })
 })

__tests__/prompt.test.ts

@@ -1,7 +1,13 @@
 import {describe, it, expect} from 'vitest'
 import * as path from 'path'
 import {fileURLToPath} from 'url'
-import {parseTemplateVariables, replaceTemplateVariables, loadPromptFile, isPromptYamlFile} from '../src/prompt'
+import {
+  parseTemplateVariables,
+  replaceTemplateVariables,
+  loadPromptFile,
+  isPromptYamlFile,
+  parseFileTemplateVariables,
+} from '../src/prompt'
 
 const __filename = fileURLToPath(import.meta.url)
 const __dirname = path.dirname(__filename)
@@ -10,8 +16,8 @@ describe('prompt.ts', () => {
   describe('parseTemplateVariables', () => {
     it('should parse simple YAML variables', () => {
       const input = `
-a: hello
+ a: hello
-b: world
+ b: world
       `
       const result = parseTemplateVariables(input)
       expect(result).toEqual({a: 'hello', b: 'world'})
@@ -19,10 +25,10 @@ b: world
 
     it('should parse multiline variables', () => {
       const input = `
-var1: hello
+ var1: hello
-var2: |
+ var2: |
-  This is a
+   This is a
-  multiline string
+   multiline string
       `
       const result = parseTemplateVariables(input)
       expect(result.var1).toBe('hello')
@@ -117,4 +123,17 @@ var2: |
       expect(() => loadPromptFile('non-existent.prompt.yml')).toThrow('Prompt file not found')
     })
   })
+
+  describe('parseFileTemplateVariables', () => {
+    it('reads file contents for variables', () => {
+      const configPath = path.join(__dirname, '../__fixtures__/prompts/json-schema.prompt.yml')
+      const data = parseFileTemplateVariables(`sample: ${configPath}`)
+      expect(data.sample).toContain('messages:')
+      expect(data.sample).toContain('responseFormat:')
+    })
+
+    it('errors on missing files', () => {
+      expect(() => parseFileTemplateVariables('x: ./does-not-exist.txt')).toThrow('was not found')
+    })
+  })
 })

action.yml

@@ -22,6 +22,10 @@ inputs:
     description: Template variables in YAML format for .prompt.yml files
     required: false
     default: ''
+  file_input:
+    description: Template variables in YAML format mapping variable names to file paths. The file contents will be used for templating.
+    required: false
+    default: ''
   model:
     description: The model to use
     required: false
@@ -50,6 +54,10 @@ inputs:
     description: Enable Model Context Protocol integration with GitHub tools
     required: false
     default: 'false'
+  github-mcp-token:
+    description: The token to use for GitHub MCP server (defaults to GITHUB_TOKEN if not specified)
+    required: false
+    default: ''
 
 # Define your outputs here.
 outputs:

package-lock.json

@@ -12,12 +12,10 @@
         "@actions/core": "^1.11.1",
         "@modelcontextprotocol/sdk": "^1.15.1",
         "js-yaml": "^4.1.0",
+        "openai": "^5.11.0",
         "pkce-challenge": "^5.0.0"
       },
       "devDependencies": {
-        "@azure-rest/ai-inference": "latest",
-        "@azure/core-auth": "latest",
-        "@azure/core-sse": "latest",
         "@eslint/compat": "^1.3.0",
         "@github/local-action": "^5.1.0",
         "@github/prettier-config": "^0.0.6",
@@ -518,44 +516,6 @@
       "integrity": "sha512-wi9JjgKLYS7U/z8PPbco+PvTb/nRWjeoFlJ1Qer83k/3C5PHQi28hiVdeE2kHXmIL99mQFawx8qt/JPjZilJ8Q==",
       "license": "MIT"
     },
-    "node_modules/@azure-rest/ai-inference": {
-      "version": "1.0.0-beta.6",
-      "resolved": "https://registry.npmjs.org/@azure-rest/ai-inference/-/ai-inference-1.0.0-beta.6.tgz",
-      "integrity": "sha512-j5FrJDTHu2P2+zwFVe5j2edasOIhqkFj+VkDjbhGkQuOoIAByF0egRkgs0G1k03HyJ7bOOT9BkRF7MIgr/afhw==",
-      "dev": true,
-      "license": "MIT",
-      "dependencies": {
-        "@azure-rest/core-client": "^2.1.0",
-        "@azure/abort-controller": "^2.1.2",
-        "@azure/core-auth": "^1.9.0",
-        "@azure/core-lro": "^2.7.2",
-        "@azure/core-rest-pipeline": "^1.18.2",
-        "@azure/core-tracing": "^1.2.0",
-        "@azure/logger": "^1.1.4",
-        "tslib": "^2.8.1"
-      },
-      "engines": {
-        "node": ">=18.0.0"
-      }
-    },
-    "node_modules/@azure-rest/core-client": {
-      "version": "2.4.0",
-      "resolved": "https://registry.npmjs.org/@azure-rest/core-client/-/core-client-2.4.0.tgz",
-      "integrity": "sha512-CjMFBcmnt0YNdRcxSSoZbtZNXudLlicdml7UrPsV03nHiWB+Bq5cu5ctieyaCuRtU7jm7+SOFtiE/g4pBFPKKA==",
-      "dev": true,
-      "license": "MIT",
-      "dependencies": {
-        "@azure/abort-controller": "^2.0.0",
-        "@azure/core-auth": "^1.3.0",
-        "@azure/core-rest-pipeline": "^1.5.0",
-        "@azure/core-tracing": "^1.0.1",
-        "@typespec/ts-http-runtime": "^0.2.2",
-        "tslib": "^2.6.2"
-      },
-      "engines": {
-        "node": ">=18.0.0"
-      }
-    },
     "node_modules/@azure/abort-controller": {
       "version": "2.1.2",
       "resolved": "https://registry.npmjs.org/@azure/abort-controller/-/abort-controller-2.1.2.tgz",
@@ -666,19 +626,6 @@
         "node": ">=18.0.0"
       }
     },
-    "node_modules/@azure/core-sse": {
-      "version": "2.3.0",
-      "resolved": "https://registry.npmjs.org/@azure/core-sse/-/core-sse-2.3.0.tgz",
-      "integrity": "sha512-jKhPpdDbVS5GlpadSKIC7V6Q4P2vEcwXi1c4CLTXs01Q/PAITES9v5J/S73+RtCMqQpsX0jGa2yPWwXi9JzdgA==",
-      "dev": true,
-      "license": "MIT",
-      "dependencies": {
-        "tslib": "^2.6.2"
-      },
-      "engines": {
-        "node": ">=20.0.0"
-      }
-    },
     "node_modules/@azure/core-tracing": {
       "version": "1.2.0",
       "resolved": "https://registry.npmjs.org/@azure/core-tracing/-/core-tracing-1.2.0.tgz",
@@ -2390,9 +2337,9 @@
       ]
     },
     "node_modules/@rollup/rollup-linux-x64-gnu": {
-      "version": "4.45.1",
+      "version": "4.46.0",
-      "resolved": "https://registry.npmjs.org/@rollup/rollup-linux-x64-gnu/-/rollup-linux-x64-gnu-4.45.1.tgz",
+      "resolved": "https://registry.npmjs.org/@rollup/rollup-linux-x64-gnu/-/rollup-linux-x64-gnu-4.46.0.tgz",
-      "integrity": "sha512-+E/lYl6qu1zqgPEnTrs4WysQtvc/Sh4fC2nByfFExqgYrqkKWp1tWIbe+ELhixnenSpBbLXNi6vbEEJ8M7fiHw==",
+      "integrity": "sha512-Cv+moII5C8RM6gZbR3cb21o6rquVDZrN2o81maROg1LFzBz2dZUwIQSxFA8GtGZ/F2KtsqQ2z3eFPBb6akvQNg==",
       "cpu": [
         "x64"
       ],
@@ -7034,6 +6981,27 @@
         "wrappy": "1"
       }
     },
+    "node_modules/openai": {
+      "version": "5.11.0",
+      "resolved": "https://registry.npmjs.org/openai/-/openai-5.11.0.tgz",
+      "integrity": "sha512-+AuTc5pVjlnTuA9zvn8rA/k+1RluPIx9AD4eDcnutv6JNwHHZxIhkFy+tmMKCvmMFDQzfA/r1ujvPWB19DQkYg==",
+      "license": "Apache-2.0",
+      "bin": {
+        "openai": "bin/cli"
+      },
+      "peerDependencies": {
+        "ws": "^8.18.0",
+        "zod": "^3.23.8"
+      },
+      "peerDependenciesMeta": {
+        "ws": {
+          "optional": true
+        },
+        "zod": {
+          "optional": true
+        }
+      }
+    },
     "node_modules/optionator": {
       "version": "0.9.4",
       "resolved": "https://registry.npmjs.org/optionator/-/optionator-0.9.4.tgz",

package.json

@@ -26,12 +26,10 @@
     "@actions/core": "^1.11.1",
     "@modelcontextprotocol/sdk": "^1.15.1",
     "js-yaml": "^4.1.0",
+    "openai": "^5.11.0",
     "pkce-challenge": "^5.0.0"
   },
   "devDependencies": {
-    "@azure-rest/ai-inference": "latest",
-    "@azure/core-auth": "latest",
-    "@azure/core-sse": "latest",
     "@eslint/compat": "^1.3.0",
     "@github/local-action": "^5.1.0",
     "@github/prettier-config": "^0.0.6",

src/helpers.ts

@@ -1,5 +1,4 @@
 import * as core from '@actions/core'
-import {GetChatCompletionsDefaultResponse} from '@azure-rest/ai-inference'
 import * as fs from 'fs'
 import {PromptConfig} from './prompt.js'
 import {InferenceRequest} from './inference.js'
@@ -29,36 +28,6 @@ export function loadContentFromFileOrInput(filePathInput: string, contentInput:
   }
 }
 
-/**
- * Helper function to handle unexpected responses from AI service
- * @param response - The response object from the AI service
- * @throws Error with appropriate error message based on response content
- */
-export function handleUnexpectedResponse(response: GetChatCompletionsDefaultResponse): never {
-  // Extract x-ms-error-code from headers if available
-  const errorCode = response.headers['x-ms-error-code']
-  const errorCodeMsg = errorCode ? ` (error code: ${errorCode})` : ''
-
-  // Check if response body exists and contains error details
-  if (response.body && response.body.error) {
-    throw response.body.error
-  }
-
-  // Handle case where response body is missing
-  if (!response.body) {
-    throw new Error(
-      `Failed to get response from AI service (status: ${response.status})${errorCodeMsg}. ` +
-        'Please check network connection and endpoint configuration.',
-    )
-  }
-
-  // Handle other error cases
-  throw new Error(
-    `AI service returned error response (status: ${response.status})${errorCodeMsg}: ` +
-      (typeof response.body === 'string' ? response.body : JSON.stringify(response.body)),
-  )
-}
-
 /**
  * Build messages array from either prompt config or legacy format
  */
@@ -66,11 +35,11 @@ export function buildMessages(
   promptConfig?: PromptConfig,
   systemPrompt?: string,
   prompt?: string,
-): Array<{role: string; content: string}> {
+): Array<{role: 'system' | 'user' | 'assistant' | 'tool'; content: string}> {
   if (promptConfig?.messages && promptConfig.messages.length > 0) {
     // Use new message format
     return promptConfig.messages.map(msg => ({
-      role: msg.role,
+      role: msg.role as 'system' | 'user' | 'assistant' | 'tool',
       content: msg.content,
     }))
   } else {

src/inference.ts

@@ -1,25 +1,16 @@
 import * as core from '@actions/core'
-import ModelClient, {isUnexpected} from '@azure-rest/ai-inference'
+import OpenAI from 'openai'
-import {AzureKeyCredential} from '@azure/core-auth'
+import {GitHubMCPClient, executeToolCalls, ToolCall} from './mcp.js'
-import {GitHubMCPClient, executeToolCalls, MCPTool, ToolCall} from './mcp.js'
-import {handleUnexpectedResponse} from './helpers.js'
 
 interface ChatMessage {
-  role: string
+  role: 'system' | 'user' | 'assistant' | 'tool'
   content: string | null
   tool_calls?: ToolCall[]
-}
+  tool_call_id?: string
-
-interface ChatCompletionsRequestBody {
-  messages: ChatMessage[]
-  max_tokens: number
-  model: string
-  response_format?: {type: 'json_schema'; json_schema: unknown}
-  tools?: MCPTool[]
 }
 
 export interface InferenceRequest {
-  messages: Array<{role: string; content: string}>
+  messages: Array<{role: 'system' | 'user' | 'assistant' | 'tool'; content: string}>
   modelName: string
   maxTokens: number
   endpoint: string
@@ -45,33 +36,38 @@ export interface InferenceResponse {
 export async function simpleInference(request: InferenceRequest): Promise<string | null> {
   core.info('Running simple inference without tools')
 
-  const client = ModelClient(request.endpoint, new AzureKeyCredential(request.token), {
+  const client = new OpenAI({
-    userAgentOptions: {userAgentPrefix: 'github-actions-ai-inference'},
+    apiKey: request.token,
+    baseURL: request.endpoint,
   })
 
-  const requestBody: ChatCompletionsRequestBody = {
+  const chatCompletionRequest: OpenAI.Chat.Completions.ChatCompletionCreateParams = {
-    messages: request.messages,
+    messages: request.messages as OpenAI.Chat.Completions.ChatCompletionMessageParam[],
     max_tokens: request.maxTokens,
     model: request.modelName,
   }
 
   // Add response format if specified
   if (request.responseFormat) {
-    requestBody.response_format = request.responseFormat
+    // eslint-disable-next-line @typescript-eslint/no-explicit-any
+    chatCompletionRequest.response_format = request.responseFormat as any
   }
 
-  const response = await client.path('/chat/completions').post({
+  try {
-    body: requestBody,
+    const response = await client.chat.completions.create(chatCompletionRequest)
-  })
 
-  if (isUnexpected(response)) {
+    if ('choices' in response) {
-    handleUnexpectedResponse(response)
+      const modelResponse = response.choices[0]?.message?.content
+      core.info(`Model response: ${modelResponse || 'No response content'}`)
+      return modelResponse || null
+    } else {
+      core.error(`Unexpected response format from API: ${JSON.stringify(response)}`)
+      return null
+    }
+  } catch (error) {
+    core.error(`API error: ${error}`)
+    throw error
   }
-
-  const modelResponse = response.body.choices[0].message.content
-  core.info(`Model response: ${modelResponse || 'No response content'}`)
-
-  return modelResponse
 }
 
 /**
@@ -83,8 +79,9 @@ export async function mcpInference(
 ): Promise<string | null> {
   core.info('Running GitHub MCP inference with tools')
 
-  const client = ModelClient(request.endpoint, new AzureKeyCredential(request.token), {
+  const client = new OpenAI({
-    userAgentOptions: {userAgentPrefix: 'github-actions-ai-inference'},
+    apiKey: request.token,
+    baseURL: request.endpoint,
   })
 
   // Start with the pre-processed messages
@@ -92,57 +89,83 @@ export async function mcpInference(
 
   let iterationCount = 0
   const maxIterations = 5 // Prevent infinite loops
+  // We want to use response_format (e.g. JSON) on the last iteration only, so the model can output
+  // the final result in the expected format without interfering with tool calls
+  let finalMessage = false
 
   while (iterationCount < maxIterations) {
     iterationCount++
     core.info(`MCP inference iteration ${iterationCount}`)
 
-    const requestBody: ChatCompletionsRequestBody = {
+    const chatCompletionRequest: OpenAI.Chat.Completions.ChatCompletionCreateParams = {
-      messages: messages,
+      messages: messages as OpenAI.Chat.Completions.ChatCompletionMessageParam[],
       max_tokens: request.maxTokens,
       model: request.modelName,
-      tools: githubMcpClient.tools,
     }
 
-    // Add response format if specified (only on first iteration to avoid conflicts)
+    // Add response format if specified (only on final iteration to avoid conflicts with tool calls)
-    if (iterationCount === 1 && request.responseFormat) {
+    if (finalMessage && request.responseFormat) {
-      requestBody.response_format = request.responseFormat
+      // eslint-disable-next-line @typescript-eslint/no-explicit-any
+      chatCompletionRequest.response_format = request.responseFormat as any
+    } else {
+      chatCompletionRequest.tools = githubMcpClient.tools as OpenAI.Chat.Completions.ChatCompletionTool[]
     }
 
-    const response = await client.path('/chat/completions').post({
+    try {
-      body: requestBody,
+      const response = await client.chat.completions.create(chatCompletionRequest)
-    })
 
-    if (isUnexpected(response)) {
+      if (!('choices' in response)) {
-      handleUnexpectedResponse(response)
+        throw new Error(`Unexpected response format from API: ${JSON.stringify(response)}`)
+      }
+
+      const assistantMessage = response.choices[0]?.message
+      const modelResponse = assistantMessage?.content
+      const toolCalls = assistantMessage?.tool_calls
+
+      core.info(`Model response: ${modelResponse || 'No response content'}`)
+
+      messages.push({
+        role: 'assistant',
+        content: modelResponse || '',
+        ...(toolCalls && {tool_calls: toolCalls as ToolCall[]}),
+      })
+
+      if (!toolCalls || toolCalls.length === 0) {
+        core.info('No tool calls requested, ending GitHub MCP inference loop')
+
+        // If we have a response format set and we haven't explicitly run one final message iteration,
+        // do another loop with the response format set
+        if (request.responseFormat && !finalMessage) {
+          core.info('Making one more MCP loop with the requested response format...')
+
+          // Add a user message requesting JSON format and try again
+          messages.push({
+            role: 'user',
+            content: `Please provide your response in the exact ${request.responseFormat.type} format specified.`,
+          })
+
+          finalMessage = true
+
+          // Continue the loop to get a properly formatted response
+          continue
+        } else {
+          return modelResponse || null
+        }
+      }
+
+      core.info(`Model requested ${toolCalls.length} tool calls`)
+
+      // Execute all tool calls via GitHub MCP
+      const toolResults = await executeToolCalls(githubMcpClient.client, toolCalls as ToolCall[])
+
+      // Add tool results to the conversation
+      messages.push(...toolResults)
+
+      core.info('Tool results added, continuing conversation...')
+    } catch (error) {
+      core.error(`OpenAI API error: ${error}`)
+      throw error
     }
-
-    const assistantMessage = response.body.choices[0].message
-    const modelResponse = assistantMessage.content
-    const toolCalls = assistantMessage.tool_calls
-
-    core.info(`Model response: ${modelResponse || 'No response content'}`)
-
-    messages.push({
-      role: 'assistant',
-      content: modelResponse || '',
-      ...(toolCalls && {tool_calls: toolCalls}),
-    })
-
-    if (!toolCalls || toolCalls.length === 0) {
-      core.info('No tool calls requested, ending GitHub MCP inference loop')
-      return modelResponse
-    }
-
-    core.info(`Model requested ${toolCalls.length} tool calls`)
-
-    // Execute all tool calls via GitHub MCP
-    const toolResults = await executeToolCalls(githubMcpClient.client, toolCalls)
-
-    // Add tool results to the conversation
-    messages.push(...toolResults)
-
-    core.info('Tool results added, continuing conversation...')
   }
 
   core.warning(`GitHub MCP inference loop exceeded maximum iterations (${maxIterations})`)

src/main.ts

@@ -5,7 +5,13 @@ import * as path from 'path'
 import {connectToGitHubMCP} from './mcp.js'
 import {simpleInference, mcpInference} from './inference.js'
 import {loadContentFromFileOrInput, buildInferenceRequest} from './helpers.js'
-import {loadPromptFile, parseTemplateVariables, isPromptYamlFile, PromptConfig} from './prompt.js'
+import {
+  loadPromptFile,
+  parseTemplateVariables,
+  isPromptYamlFile,
+  PromptConfig,
+  parseFileTemplateVariables,
+} from './prompt.js'
 
 const RESPONSE_FILE = 'modelResponse.txt'
 
@@ -18,6 +24,7 @@ export async function run(): Promise<void> {
   try {
     const promptFilePath = core.getInput('prompt-file')
     const inputVariables = core.getInput('input')
+    const fileInputVariables = core.getInput('file_input')
 
     let promptConfig: PromptConfig | undefined = undefined
     let systemPrompt: string | undefined = undefined
@@ -27,8 +34,10 @@ export async function run(): Promise<void> {
     if (promptFilePath && isPromptYamlFile(promptFilePath)) {
       core.info('Using prompt YAML file format')
 
-      // Parse template variables
+      // Parse template variables from both string inputs and file-based inputs
-      const templateVariables = parseTemplateVariables(inputVariables)
+      const stringVars = parseTemplateVariables(inputVariables)
+      const fileVars = parseFileTemplateVariables(fileInputVariables)
+      const templateVariables = {...stringVars, ...fileVars}
 
       // Load and process prompt file
       promptConfig = loadPromptFile(promptFilePath, templateVariables)
@@ -49,6 +58,9 @@ export async function run(): Promise<void> {
       throw new Error('GITHUB_TOKEN is not set')
     }
 
+    // Get GitHub MCP token (use dedicated token if provided, otherwise fall back to main token)
+    const githubMcpToken = core.getInput('github-mcp-token') || token
+
     const endpoint = core.getInput('endpoint')
 
     // Build the inference request with pre-processed messages and response format
@@ -67,7 +79,7 @@ export async function run(): Promise<void> {
     let modelResponse: string | null = null
 
     if (enableMcp) {
-      const mcpClient = await connectToGitHubMCP(inferenceRequest.token)
+      const mcpClient = await connectToGitHubMCP(githubMcpToken)
 
       if (mcpClient) {
         modelResponse = await mcpInference(inferenceRequest, mcpClient)
@@ -91,9 +103,14 @@ export async function run(): Promise<void> {
     if (error instanceof Error) {
       core.setFailed(error.message)
     } else {
-      core.setFailed('An unexpected error occurred')
+      core.setFailed(`An unexpected error occurred: ${JSON.stringify(error, null, 2)}`)
     }
+    // Force exit to prevent hanging on open connections
+    process.exit(1)
   }
+
+  // Force exit to prevent hanging on open connections
+  process.exit(0)
 }
 
 function tempDir(): string {

src/prompt.ts

@@ -37,6 +37,47 @@ export function parseTemplateVariables(input: string): TemplateVariables {
   }
 }
 
+/**
+ * Parse file-based template variables from YAML input string. The YAML should map
+ * variable names to file paths. File contents are read and returned as variables.
+ */
+export function parseFileTemplateVariables(fileInput: string): TemplateVariables {
+  if (!fileInput.trim()) {
+    return {}
+  }
+
+  try {
+    const parsed = yaml.load(fileInput) as Record<string, unknown>
+    if (typeof parsed !== 'object' || parsed === null) {
+      throw new Error('File template variables must be a YAML object')
+    }
+
+    const result: TemplateVariables = {}
+    for (const [key, value] of Object.entries(parsed)) {
+      if (typeof value !== 'string') {
+        throw new Error(`File template variable '${key}' must be a string file path`)
+      }
+      const filePath = value
+      if (!fs.existsSync(filePath)) {
+        throw new Error(`File for template variable '${key}' was not found: ${filePath}`)
+      }
+      try {
+        result[key] = fs.readFileSync(filePath, 'utf-8')
+      } catch (err) {
+        throw new Error(
+          `Failed to read file for template variable '${key}' at path '${filePath}': ${err instanceof Error ? err.message : 'Unknown error'}`,
+        )
+      }
+    }
+
+    return result
+  } catch (error) {
+    throw new Error(
+      `Failed to parse file template variables: ${error instanceof Error ? error.message : 'Unknown error'}`,
+    )
+  }
+}
+
 /**
  * Replace template variables in text using {{variable}} syntax
  */