|
| 1 | +name: Q&A evaluation (python_server) |
| 2 | + |
| 3 | +on: |
| 4 | + push: |
| 5 | + paths: |
| 6 | + - "python_server/**" |
| 7 | + - ".github/workflows/qa-eval.yml" |
| 8 | + - ".github/actions/**" |
| 9 | + |
| 10 | +concurrency: |
| 11 | + group: ${{ github.workflow }}-${{ github.ref }} |
| 12 | + cancel-in-progress: true |
| 13 | + |
| 14 | +env: |
| 15 | + DEFAULT_WORKING_DIRECTORY: python_server |
| 16 | + YOUTUBE_API_KEYS: '[]' |
| 17 | + GOOGLE_APPLICATION_CREDENTIALS: "/tmp" |
| 18 | + GOOGLE_DRIVE_FOLDER_ID: "" |
| 19 | + # PostgreSQL の情報。使用してないが Pythontic settings がコケるため設定しておく |
| 20 | + PG_HOST: localhost |
| 21 | + PG_PORT: 5432 |
| 22 | + PG_USER: postgres |
| 23 | + PG_PASSWORD: password |
| 24 | + PG_DATABASE: aituber_dev |
| 25 | + |
| 26 | + |
| 27 | +defaults: |
| 28 | + run: |
| 29 | + working-directory: python_server |
| 30 | + |
| 31 | +jobs: |
| 32 | + eval: |
| 33 | + runs-on: ubuntu-latest |
| 34 | + |
| 35 | + permissions: |
| 36 | + contents: read |
| 37 | + pull-requests: write # to post comments |
| 38 | + |
| 39 | + steps: |
| 40 | + - uses: actions/checkout@v4 |
| 41 | + |
| 42 | + - name: setup Python & Poetry |
| 43 | + uses: ./.github/actions/setup-python-and-poetry |
| 44 | + with: |
| 45 | + working-directory: ${{ env.DEFAULT_WORKING_DIRECTORY }} |
| 46 | + |
| 47 | + - name: Save FAISS knowledge database |
| 48 | + run: poetry run python -m src.cli.save_faiss_knowledge_db |
| 49 | + env: |
| 50 | + OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }} |
| 51 | + YOUTUBE_API_KEY: ${{ secrets.YOUTUBE_API_KEY }} |
| 52 | + YT_ID: ${{ secrets.YT_ID }} |
| 53 | + ELEVENLABS_API_KEY: ${{ secrets.ELEVENLABS_API_KEY }} |
| 54 | + AZURE_SPEECH_KEY: ${{ secrets.AZURE_SPEECH_KEY }} |
| 55 | + |
| 56 | + - name: Save FAISS database |
| 57 | + run: poetry run python -m src.cli.save_faiss_db --for-eval |
| 58 | + env: |
| 59 | + OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }} |
| 60 | + YOUTUBE_API_KEY: ${{ secrets.YOUTUBE_API_KEY }} |
| 61 | + YT_ID: ${{ secrets.YT_ID }} |
| 62 | + ELEVENLABS_API_KEY: ${{ secrets.ELEVENLABS_API_KEY }} |
| 63 | + AZURE_SPEECH_KEY: ${{ secrets.AZURE_SPEECH_KEY }} |
| 64 | + |
| 65 | + - name: Run evaluation |
| 66 | + run: poetry run python -m src.cli.rag_evaluation.evaluate --output-path result.md |
| 67 | + env: |
| 68 | + OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }} |
| 69 | + YOUTUBE_API_KEY: ${{ secrets.YOUTUBE_API_KEY }} |
| 70 | + YT_ID: ${{ secrets.YT_ID }} |
| 71 | + ELEVENLABS_API_KEY: ${{ secrets.ELEVENLABS_API_KEY }} |
| 72 | + AZURE_SPEECH_KEY: ${{ secrets.AZURE_SPEECH_KEY }} |
| 73 | + |
| 74 | + - name: Run evaluation with cosine-similarity-based document retrieval |
| 75 | + run: poetry run python -m src.cli.rag_evaluation.evaluate --output-path result_cos_sim.md --doc-retrieval-type cosine |
| 76 | + env: |
| 77 | + OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }} |
| 78 | + YOUTUBE_API_KEY: ${{ secrets.YOUTUBE_API_KEY }} |
| 79 | + YT_ID: ${{ secrets.YT_ID }} |
| 80 | + ELEVENLABS_API_KEY: ${{ secrets.ELEVENLABS_API_KEY }} |
| 81 | + AZURE_SPEECH_KEY: ${{ secrets.AZURE_SPEECH_KEY }} |
| 82 | + |
| 83 | + - name: Show contents in result.md |
| 84 | + run: cat result.md |
| 85 | + |
| 86 | + - name: Post a message to Pull Request threads |
| 87 | + uses: actions/github-script@v7 |
| 88 | + env: |
| 89 | + RESULT_MD: result.md |
| 90 | + RESULT_COSINE: result_cos_sim.md |
| 91 | + with: |
| 92 | + script: | |
| 93 | + process.chdir(process.env.DEFAULT_WORKING_DIRECTORY); |
| 94 | +
|
| 95 | + const fs = require("fs"); |
| 96 | + const resultDefault = fs.readFileSync(process.env.RESULT_MD, "utf8"); |
| 97 | + const resultCosSim = fs.readFileSync(process.env.RESULT_COSINE, "utf8"); |
| 98 | +
|
| 99 | + let commentBody = "" |
| 100 | + commentBody += "### Evaluation results with the default document retrieval type\n\n" |
| 101 | + commentBody += resultDefault |
| 102 | + commentBody += "\n\n" |
| 103 | + commentBody += "### Evaluation results with the cosine-similarity-based document retrieval\n\n" |
| 104 | + commentBody += "<details><summary>Click to expand</summary>\n\n" |
| 105 | + commentBody += resultCosSim |
| 106 | + commentBody += "</details>" |
| 107 | +
|
| 108 | + async function getTargetPullRequests() { |
| 109 | + if (context.eventName === "pull_request") { |
| 110 | + return [context.payload.pull_request]; |
| 111 | + } |
| 112 | + if (context.eventName === "push") { |
| 113 | + const pullRequests = ( |
| 114 | + await github.rest.repos.listPullRequestsAssociatedWithCommit({ |
| 115 | + owner: context.repo.owner, |
| 116 | + repo: context.repo.repo, |
| 117 | + commit_sha: context.sha, |
| 118 | + }) |
| 119 | + ).data; |
| 120 | + const openPullRequests = pullRequests.filter(pr => pr.state === "open") |
| 121 | + return openPullRequests; |
| 122 | + } |
| 123 | + throw new Error(`Unsupported event type: ${context.eventName}`); |
| 124 | + } |
| 125 | + const targetPullRequests = await getTargetPullRequests(); |
| 126 | +
|
| 127 | + core.startGroup('Found PRs:') |
| 128 | + console.log(targetPullRequests) |
| 129 | + core.endGroup() |
| 130 | +
|
| 131 | + for (const pr of targetPullRequests) { |
| 132 | + const title = `## Eval results on [\`${context.sha.slice(0, 7)}\`](${pr.html_url}/commits/${context.sha})\n\n`; |
| 133 | + const commentText = title + commentBody; |
| 134 | +
|
| 135 | + await github.rest.issues.createComment({ |
| 136 | + owner: context.repo.owner, |
| 137 | + repo: context.repo.repo, |
| 138 | + issue_number: pr.number, |
| 139 | + body: commentText, |
| 140 | + }); |
| 141 | + core.info(`Created comment on PR #${pr.number}`); |
| 142 | + } |
0 commit comments