actions/ai-inference
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat: Add system-prompt-file input for file-based system prompts

Browse Source 
This enhancement adds the ability to load a system prompt from a file, similar to
the existing prompt-file functionality, providing more flexibility when working with
complex system prompts.

Key changes:
- Added new `system-prompt-file` input to action.yml with proper description
- Updated main.ts implementation to handle file-based system prompts with:
  - File existence checking and appropriate error handling
  - Proper precedence (system-prompt-file takes priority over system-prompt)
  - Consistent error messages with existing prompt-file implementation

Test coverage added:
- Basic usage: Test verifies system-prompt-file loads content correctly
- Error handling: Test ensures proper errors when system-prompt-file doesn't exist
- Precedence: Test confirms system-prompt-file overrides system-prompt when both exist
- Integration: Test validates both prompt-file and system-prompt-file work together
- Max tokens: Test verifies custom token limits are properly passed to the model
- Testing infrastructure: Improved mock implementations that detect unexpected calls

Documentation:
- Updated README.md with system-prompt-file in inputs table
- Added dedicated usage example for system-prompt-file
- Fixed formatting in inputs table

CI/CD:
- Updated workflow to test system-prompt-file functionality with actual file content

This feature maintains backward compatibility while adding a useful option
for workflows that need to use more complex system prompts stored in files.
This commit is contained in:
Matthew Leibowitz
2025-05-24 03:50:15 +02:00
parent f8ee4c952b
commit 91ba53d8b4
8 changed files with 381 additions and 22 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
.github/workflows/ci.yml vendored
View File

@@ -80,11 +80,16 @@ jobs:
- name: Create Prompt File
run: echo "hello" > prompt.txt
- name: Create System Prompt File
run:
echo "You are a helpful AI assistant for testing." > system-prompt.txt
- name: Test Local Action with Prompt File
id: test-action-prompt-file
uses: ./
with:
prompt-file: prompt.txt
system-prompt-file: system-prompt.txt
env:
GITHUB_TOKEN: ${{ github.token }}

34
README.md
View File

@@ -47,6 +47,21 @@ steps:
prompt-file: './path/to/prompt.txt'
```
### Using a system prompt file
In addition to the regular prompt, you can provide a system prompt file instead
of an inline system prompt:
```yaml
steps:
- name: Run AI Inference with System Prompt File
id: inference
uses: actions/ai-inference@v1
with:
prompt: 'Hello!'
system-prompt-file: './path/to/system-prompt.txt'
```
### Read output from file instead of output
This can be useful when model response exceeds actions output limit
@@ -70,15 +85,16 @@ steps:
Various inputs are defined in [`action.yml`](action.yml) to let you configure
the action:
| Name | Description | Default |
| --------------- | ------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------ |
| `token` | Token to use for inference. Typically the GITHUB_TOKEN secret | `github.token` |
| `prompt` | The prompt to send to the model | N/A |
| `prompt-file` | Path to a file containing the prompt. If both `prompt` and `prompt-file` are provided, `prompt-file` takes precedence | `""` |
| `system-prompt` | The system prompt to send to the model | `""` |
| `model` | The model to use for inference. Must be available in the [GitHub Models](https://github.com/marketplace?type=models) catalog | `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 |
| Name | Description | Default |
| -------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------ |
| `token` | Token to use for inference. Typically the GITHUB_TOKEN secret | `github.token` |
| `prompt` | The prompt to send to the model | N/A |
| `prompt-file` | Path to a file containing the prompt. If both `prompt` and `prompt-file` are provided, `prompt-file` takes precedence | `""` |
| `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 | `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 |
## Outputs

318
__tests__/main.test.ts
View File

@@ -28,8 +28,13 @@ jest.unstable_mockModule('@azure-rest/ai-inference', () => ({
isUnexpected: jest.fn(() => false)
}))
const mockExistsSync = jest.fn().mockReturnValue(true)
const mockReadFileSync = jest.fn().mockReturnValue('Hello, AI!')
// Default to throwing errors to catch unexpected calls
const mockExistsSync = jest.fn().mockImplementation(() => {
throw new Error('Unexpected call to existsSync - test should override this implementation')
})
const mockReadFileSync = jest.fn().mockImplementation(() => {
throw new Error('Unexpected call to readFileSync - test should override this implementation')
})
jest.unstable_mockModule('fs', () => ({
existsSync: mockExistsSync,
@@ -43,12 +48,17 @@ jest.unstable_mockModule('@actions/core', () => core)
const { run } = await import('../src/main.js')
describe('main.ts', () => {
// Reset all mocks before each test
beforeEach(() => {
jest.clearAllMocks();
});
it('Sets the response output', async () => {
// Set the action's inputs as return values from core.getInput().
core.getInput.mockImplementation((name) => {
if (name === 'prompt') return 'Hello, AI!'
if (name === 'system_prompt') return 'You are a test assistant.'
if (name === 'model_name') return 'gpt-4o'
if (name === 'system-prompt') return 'You are a test assistant.'
if (name === 'token') return 'fake-token'
return ''
})
@@ -67,11 +77,12 @@ describe('main.ts', () => {
)
})
it('Sets a failed status', async () => {
it('Sets a failed status when no prompt is set', async () => {
// Clear the getInput mock and simulate no prompt or prompt-file input
core.getInput.mockImplementation((name) => {
if (name === 'prompt') return ''
if (name === 'prompt_file') return ''
if (name === 'prompt-file') return ''
if (name === 'token') return 'fake-token'
return ''
})
@@ -83,10 +94,21 @@ describe('main.ts', () => {
it('uses prompt-file', async () => {
const promptFile = 'prompt.txt'
const promptContent = 'This is a prompt from a file'
// Set up mock to return specific content for the prompt file
mockExistsSync.mockReturnValue(true)
mockReadFileSync.mockImplementation((path, encoding) => {
if (path === promptFile && encoding === 'utf-8') {
return promptContent
}
throw new Error(`Unexpected file read: ${path}`)
})
core.getInput.mockImplementation((name) => {
if (name === 'prompt-file') return promptFile
if (name === 'system-prompt') return 'You are a test assistant.'
if (name === 'model-name') return 'gpt-4o'
if (name === 'token') return 'fake-token'
return ''
})
@@ -105,4 +127,286 @@ describe('main.ts', () => {
expect.stringContaining('modelResponse.txt')
)
})
it('handles non-existent prompt-file with an error', async () => {
const promptFile = 'non-existent-prompt.txt'
// Mock the file not existing
mockExistsSync.mockImplementation((path) => {
if (path === promptFile) {
return false
}
return true
})
core.getInput.mockImplementation((name) => {
if (name === 'prompt-file') return promptFile
if (name === 'token') return 'fake-token'
return ''
})
await run()
// Verify that the error was correctly reported
expect(core.setFailed).toHaveBeenCalledWith(
`Prompt file not found: ${promptFile}`
)
})
it('prefers prompt-file over prompt when both are provided', async () => {
const promptFile = 'prompt.txt'
const promptFileContent = 'This is a prompt from a file that should be used'
const promptString = 'This is a direct prompt that should be ignored'
// Set up mock to return specific content for the prompt file
mockExistsSync.mockReturnValue(true)
mockReadFileSync.mockImplementation((path, encoding) => {
if (path === promptFile && encoding === 'utf-8') {
return promptFileContent
}
throw new Error(`Unexpected file read: ${path}`)
})
core.getInput.mockImplementation((name) => {
if (name === 'prompt') return promptString
if (name === 'prompt-file') return promptFile
if (name === 'system-prompt') return 'You are a test assistant.'
if (name === 'token') return 'fake-token'
return ''
})
await run()
expect(mockExistsSync).toHaveBeenCalledWith(promptFile)
expect(mockReadFileSync).toHaveBeenCalledWith(promptFile, 'utf-8')
// Check that the post call was made with the prompt from the file, not the input parameter
expect(mockPost).toHaveBeenCalledWith({
body: {
messages: [
{
role: 'system',
content: expect.any(String)
},
{ role: 'user', content: promptFileContent } // Should use the file content, not the string input
],
max_tokens: expect.any(Number),
model: expect.any(String)
}
})
expect(core.setOutput).toHaveBeenNthCalledWith(
1,
'response',
'Hello, user!'
)
expect(core.setOutput).toHaveBeenNthCalledWith(
2,
'response-file',
expect.stringContaining('modelResponse.txt')
)
})
it('uses system-prompt-file', async () => {
const systemPromptFile = 'system-prompt.txt'
const systemPromptContent = 'You are a specialized system assistant for testing'
// Set up mock to return specific content for the system prompt file
mockExistsSync.mockReturnValue(true)
mockReadFileSync.mockImplementation((path, encoding) => {
if (path === systemPromptFile && encoding === 'utf-8') {
return systemPromptContent
}
throw new Error(`Unexpected file read: ${path}`)
})
core.getInput.mockImplementation((name) => {
if (name === 'prompt') return 'Hello, AI!'
if (name === 'system-prompt-file') return systemPromptFile
if (name === 'token') return 'fake-token'
return ''
})
await run()
expect(mockExistsSync).toHaveBeenCalledWith(systemPromptFile)
expect(mockReadFileSync).toHaveBeenCalledWith(systemPromptFile, 'utf-8')
expect(core.setOutput).toHaveBeenNthCalledWith(
1,
'response',
'Hello, user!'
)
expect(core.setOutput).toHaveBeenNthCalledWith(
2,
'response-file',
expect.stringContaining('modelResponse.txt')
)
})
it('handles non-existent system-prompt-file with an error', async () => {
const systemPromptFile = 'non-existent-system-prompt.txt'
// Mock the file not existing
mockExistsSync.mockImplementation((path) => {
if (path === systemPromptFile) {
return false
}
return true
})
core.getInput.mockImplementation((name) => {
if (name === 'prompt') return 'Hello, AI!'
if (name === 'system-prompt-file') return systemPromptFile
if (name === 'token') return 'fake-token'
return ''
})
await run()
// Verify that the error was correctly reported
expect(core.setFailed).toHaveBeenCalledWith(
`System prompt file not found: ${systemPromptFile}`
)
})
it('prefers system-prompt-file over system-prompt when both are provided', async () => {
const systemPromptFile = 'system-prompt.txt'
const systemPromptFileContent = 'You are a specialized system assistant from file'
const systemPromptString = 'You are a basic system assistant from input parameter'
// Set up mock to return specific content for the system prompt file
mockExistsSync.mockReturnValue(true)
mockReadFileSync.mockImplementation((path, encoding) => {
if (path === systemPromptFile && encoding === 'utf-8') {
return systemPromptFileContent
}
throw new Error(`Unexpected file read: ${path}`)
})
core.getInput.mockImplementation((name) => {
if (name === 'prompt') return 'Hello, AI!'
if (name === 'system-prompt-file') return systemPromptFile
if (name === 'system-prompt') return systemPromptString
if (name === 'token') return 'fake-token'
return ''
})
await run()
expect(mockExistsSync).toHaveBeenCalledWith(systemPromptFile)
expect(mockReadFileSync).toHaveBeenCalledWith(systemPromptFile, 'utf-8')
// Check that the post call was made with the system prompt from the file, not the input parameter
expect(mockPost).toHaveBeenCalledWith({
body: {
messages: [
{
role: 'system',
content: systemPromptFileContent // Should use the file content, not the string input
},
{ role: 'user', content: 'Hello, AI!' }
],
max_tokens: expect.any(Number),
model: expect.any(String)
}
})
expect(core.setOutput).toHaveBeenNthCalledWith(
1,
'response',
'Hello, user!'
)
expect(core.setOutput).toHaveBeenNthCalledWith(
2,
'response-file',
expect.stringContaining('modelResponse.txt')
)
})
it('uses both prompt-file and system-prompt-file together', async () => {
const promptFile = 'prompt.txt'
const promptContent = 'This is a prompt from a file'
const systemPromptFile = 'system-prompt.txt'
const systemPromptContent = 'You are a specialized system assistant from file'
// Set up mock to return specific content for both files
mockExistsSync.mockReturnValue(true)
mockReadFileSync.mockImplementation((path, encoding) => {
if (path === promptFile && encoding === 'utf-8') {
return promptContent
} else if (path === systemPromptFile && encoding === 'utf-8') {
return systemPromptContent
}
throw new Error(`Unexpected file read: ${path}`)
})
core.getInput.mockImplementation((name) => {
if (name === 'prompt-file') return promptFile
if (name === 'system-prompt-file') return systemPromptFile
if (name === 'token') return 'fake-token'
return ''
})
await run()
expect(mockExistsSync).toHaveBeenCalledWith(promptFile)
expect(mockExistsSync).toHaveBeenCalledWith(systemPromptFile)
expect(mockReadFileSync).toHaveBeenCalledWith(promptFile, 'utf-8')
expect(mockReadFileSync).toHaveBeenCalledWith(systemPromptFile, 'utf-8')
// Check that the post call was made with both the prompt and system prompt from files
expect(mockPost).toHaveBeenCalledWith({
body: {
messages: [
{
role: 'system',
content: systemPromptContent
},
{ role: 'user', content: promptContent }
],
max_tokens: expect.any(Number),
model: expect.any(String)
}
})
expect(core.setOutput).toHaveBeenNthCalledWith(
1,
'response',
'Hello, user!'
)
expect(core.setOutput).toHaveBeenNthCalledWith(
2,
'response-file',
expect.stringContaining('modelResponse.txt')
)
})
it('passes custom max-tokens parameter to the model', async () => {
const customMaxTokens = 500
core.getInput.mockImplementation((name) => {
if (name === 'prompt') return 'Hello, AI!'
if (name === 'system-prompt') return 'You are a test assistant.'
if (name === 'token') return 'fake-token'
if (name === 'max-tokens') return customMaxTokens.toString()
return ''
})
await run()
// Check that the post call was made with the correct max_tokens parameter
expect(mockPost).toHaveBeenCalledWith({
body: {
messages: expect.any(Array),
max_tokens: customMaxTokens,
model: expect.any(String)
}
})
expect(core.setOutput).toHaveBeenNthCalledWith(
1,
'response',
'Hello, user!'
)
})
})

4
action.yml
View File

@@ -29,6 +29,10 @@ inputs:
description: The system prompt for the model
required: false
default: 'You are a helpful assistant'
system-prompt-file:
description: Path to a file containing the system prompt
required: false
default: ''
max-tokens:
description: The maximum number of tokens to generate
required: false

17
dist/index.js generated vendored
View File

@@ -33574,7 +33574,22 @@ async function run() {
else {
throw new Error('prompt is not set');
}
const systemPrompt = coreExports.getInput('system-prompt');
const systemPromptFile = coreExports.getInput('system-prompt-file');
const systemPromptString = coreExports.getInput('system-prompt');
let systemPrompt;
if (systemPromptFile !== undefined && systemPromptFile !== '') {
if (!fs.existsSync(systemPromptFile)) {
throw new Error(`System prompt file not found: ${systemPromptFile}`);
}
systemPrompt = fs.readFileSync(systemPromptFile, 'utf-8');
}
else if (systemPromptString !== undefined && systemPromptString !== '') {
systemPrompt = systemPromptString;
}
else {
// Use default system prompt
systemPrompt = 'You are a helpful assistant';
}
const modelName = coreExports.getInput('model');
const maxTokens = parseInt(coreExports.getInput('max-tokens'), 10);
const token = coreExports.getInput('token') || process.env['GITHUB_TOKEN'];

2
dist/index.js.map generated vendored
View File

File diff suppressed because one or more lines are too long

6
package-lock.json generated
View File

@@ -12,9 +12,9 @@
"@actions/core": "^1.11.1"
},
"devDependencies": {
"@azure-rest/ai-inference": "*",
"@azure/core-auth": "*",
"@azure/core-sse": "*",
"@azure-rest/ai-inference": "latest",
"@azure/core-auth": "latest",
"@azure/core-sse": "latest",
"@eslint/compat": "^1.2.8",
"@github/local-action": "^3.2.0",
"@jest/globals": "^29.7.0",

17
src/main.ts
View File

@@ -29,7 +29,22 @@ export async function run(): Promise<void> {
throw new Error('prompt is not set')
}
const systemPrompt: string = core.getInput('system-prompt')
const systemPromptFile: string = core.getInput('system-prompt-file')
const systemPromptString: string = core.getInput('system-prompt')
let systemPrompt: string
if (systemPromptFile !== undefined && systemPromptFile !== '') {
if (!fs.existsSync(systemPromptFile)) {
throw new Error(`System prompt file not found: ${systemPromptFile}`)
}
systemPrompt = fs.readFileSync(systemPromptFile, 'utf-8')
} else if (systemPromptString !== undefined && systemPromptString !== '') {
systemPrompt = systemPromptString
} else {
// Use default system prompt
systemPrompt = 'You are a helpful assistant'
}
const modelName: string = core.getInput('model')
const maxTokens: number = parseInt(core.getInput('max-tokens'), 10)