Skip to content

Commit a1ba000

Browse files
MichaelFisher1997MichaelFisher1997
committed
feat(workflows): add previous reviews tracking to opencode-pr workflows
- Add step to fetch previous automated reviews from opencode-agent[bot] - Update prompts to include previous reviews context - Add instructions to check and acknowledge fixed issues - Prevent duplicate reports by only reporting new/unresolved issues - Align Nix variant with standard version structure using GitHub Action - Keep ubuntu-latest runner for both variants
1 parent 95b21c1 commit a1ba000

2 files changed

Lines changed: 232 additions & 76 deletions

File tree

‎workflows/opencode/pr/opencode-pr-nix.yml‎

Lines changed: 162 additions & 66 deletions
Original file line numberDiff line numberDiff line change
@@ -22,94 +22,190 @@ on:
2222
types: [opened, synchronize, reopened, ready_for_review]
2323

2424
jobs:
25-
review:
25+
opencode:
26+
# Don't run on draft PRs; do run when they become ready_for_review.
27+
if: ${{ github.event.pull_request.draft == false }}
2628
runs-on: ubuntu-latest
27-
if: github.event.pull_request.draft == false
28-
29+
timeout-minutes: 10
30+
permissions:
31+
id-token: write
32+
contents: write
33+
pull-requests: write
34+
issues: read
35+
2936
steps:
30-
- name: Checkout code
37+
- name: Checkout repository
3138
uses: actions/checkout@v4
3239
with:
3340
fetch-depth: 0
3441

42+
- name: Configure git
43+
run: |
44+
git config user.name "github-actions[bot]"
45+
git config user.email "github-actions[bot]@users.noreply.github.com"
46+
3547
- name: Install Nix
3648
uses: DeterminateSystems/nix-installer-action@v16
3749

38-
- name: Cache Nix store
50+
- name: Cache Nix Store
3951
uses: nix-community/cache-nix-action@v7
4052
with:
4153
primary-key: nix-${{ runner.os }}-${{ hashFiles('flake.nix', 'flake.lock') }}
4254
restore-prefixes-first-match: nix-${{ runner.os }}-
43-
gc-max-store-size: 1G
44-
purge: true
45-
purge-prefixes: nix-${{ runner.os }}-
46-
purge-created: 0
47-
purge-primary-key: never
48-
49-
- name: Get PR diff
50-
id: diff
55+
56+
- name: Fetch previous opencode reviews
57+
id: previous-reviews
5158
run: |
5259
nix develop --command bash -c '
53-
git diff origin/${{ github.base_ref }}...HEAD > pr_diff.txt
60+
# Get PR number
61+
PR_NUMBER="${{ github.event.pull_request.number }}"
62+
63+
# Fetch review comments from opencode-agent
64+
echo "Fetching previous automated reviews..."
65+
66+
# Get all reviews
67+
gh api /repos/${{ github.repository }}/pulls/${PR_NUMBER}/reviews \
68+
--jq ".[] | select(.user.login == \\"opencode-agent[bot]\\") | {body: .body, submitted_at: .submitted_at}" > /tmp/opencode_reviews.json 2>/dev/null || echo "[]" > /tmp/opencode_reviews.json
69+
70+
# Get all PR review comments (inline comments)
71+
gh api /repos/${{ github.repository }}/pulls/${PR_NUMBER}/comments \
72+
--jq ".[] | select(.user.login == \\"opencode-agent[bot]\\") | {body: .body, path: .path, line: .line, created_at: .created_at}" > /tmp/opencode_comments.json 2>/dev/null || echo "[]" > /tmp/opencode_comments.json
73+
74+
# Format the previous reviews for the prompt
75+
REVIEW_CONTENT="## Previous Automated Reviews from opencode-agent:\n\n"
76+
77+
# Process reviews
78+
if [ -s /tmp/opencode_reviews.json ] && [ "$(cat /tmp/opencode_reviews.json)" != "[]" ]; then
79+
while IFS= read -r review; do
80+
if [ -n "$review" ] && [ "$review" != "null" ]; then
81+
body=$(echo "$review" | jq -r ".body // empty")
82+
date=$(echo "$review" | jq -r ".submitted_at // empty")
83+
if [ -n "$body" ] && [ "$body" != "null" ]; then
84+
REVIEW_CONTENT="${REVIEW_CONTENT}### Review from ${date}\n${body}\n\n---\n\n"
85+
fi
86+
fi
87+
done < /tmp/opencode_reviews.json
88+
else
89+
REVIEW_CONTENT="${REVIEW_CONTENT}No previous automated reviews found.\n"
90+
fi
91+
92+
# Write to environment file
93+
echo "PREVIOUS_REVIEWS<<EOF" >> "$GITHUB_ENV"
94+
printf "%s\n" "$REVIEW_CONTENT" >> "$GITHUB_ENV"
95+
echo "EOF" >> "$GITHUB_ENV"
96+
97+
echo "Previous reviews fetched and formatted for context"
5498
'
55-
echo "diff_path=pr_diff.txt" >> $GITHUB_OUTPUT
99+
env:
100+
GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
56101

57-
- name: Get previous review comments
58-
id: previous-comments
59-
uses: actions/github-script@v7
60-
with:
61-
script: |
62-
const comments = await github.rest.issues.listComments({
63-
owner: context.repo.owner,
64-
repo: context.repo.repo,
65-
issue_number: context.issue.number
66-
});
67-
const opencodeComments = comments.data.filter(c =>
68-
c.user.login === 'github-actions[bot]' &&
69-
c.body.includes('OpenCode AI Review')
70-
);
71-
return opencodeComments.map(c => c.body).join('\n---\n');
72-
73-
- name: Run OpenCode AI Review
102+
- name: Run opencode
103+
uses: anomalyco/opencode/github@latest
74104
env:
75105
KIMI_API_KEY: ${{ secrets.KIMI_API_KEY }}
76-
PR_TITLE: ${{ github.event.pull_request.title }}
77-
PR_DESCRIPTION: ${{ github.event.pull_request.body }}
78-
PREVIOUS_COMMENTS: ${{ steps.previous-comments.outputs.result }}
79-
run: |
80-
nix develop --command bash -c '
81-
DIFF_CONTENT=$(cat pr_diff.txt)
106+
with:
107+
model: kimi-for-coding/k2p5
108+
prompt: |
109+
You are reviewing a pull request.
110+
111+
${{ env.PREVIOUS_REVIEWS }}
112+
113+
---
114+
115+
**YOUR TASK:** Analyze the CURRENT code changes and previous reviews above, then output your review in the following STRICT STRUCTURE:
116+
117+
**CRITICAL INSTRUCTIONS:**
118+
1. **CHECK PREVIOUS ISSUES FIRST:** Look at the "Previous Automated Reviews" section above. For each issue previously reported (Critical, High, Medium, Low), verify if it still exists in the current code.
119+
2. **ACKNOWLEDGE FIXES:** If a previously reported issue has been fixed, state "✅ **[FIXED]** Previous issue: [brief description]" in the appropriate section.
120+
3. **ONLY REPORT NEW/UNRESOLVED ISSUES:** Do NOT re-report issues that have already been fixed. Only report issues that are still present in the current code.
121+
4. **TRACK CHANGES:** If an issue was reported in a previous review but the code has changed, verify the new code and report the issue with updated file:line references if it still exists.
122+
123+
---
124+
125+
## 📋 Summary
126+
First, check if the PR description mentions any linked issues (e.g., "Closes #123", "Fixes #456", "Resolves #789").
82127
83-
# Build prompt using printf and jq
84-
PROMPT=$(printf "%s\n\nTitle: %s\n\nDescription: %s\n\nPrevious Review Context:\n%s\n\nCode Changes:\n%s\n\nProvide a structured review with:\n1. Overall assessment (merge-ready, needs changes, etc.)\n2. Critical issues (if any)\n3. High-priority improvements\n4. Medium-priority suggestions\n5. Low-priority nitpicks\n6. Confidence score (0-100)\n\nFormat as markdown with clear headers." \
85-
"You are an expert code reviewer. Review the following pull request:" \
86-
"$PR_TITLE" \
87-
"$PR_DESCRIPTION" \
88-
"$PREVIOUS_COMMENTS" \
89-
"$DIFF_CONTENT")
128+
If linked issues are found:
129+
- Mention the issue number(s) explicitly
130+
- Verify the PR actually implements what the issue(s) requested
131+
- State whether the implementation fully satisfies the issue requirements
90132
91-
# Create JSON payload using jq
92-
jq -n --arg content "$PROMPT" "{\n model: \\"kimi-latest\\",\n messages: [{\n role: \\"user\\",\n content: $content\n }]\n }" > prompt.json
133+
Then provide 2-3 sentences summarizing the PR purpose, scope, and overall quality.
134+
135+
## 🔴 Critical Issues (Must Fix - Blocks Merge)
136+
**IMPORTANT:** Check previous reviews first. If critical issues were reported before, verify if they're fixed. If fixed, say "✅ All previously reported critical issues have been resolved."
93137
94-
curl -X POST https://api.opencode.ai/v1/chat/completions \
95-
-H "Authorization: Bearer $KIMI_API_KEY" \
96-
-H "Content-Type: application/json" \
97-
-d @prompt.json > review_response.json
138+
Only report NEW critical issues that could cause crashes, security vulnerabilities, data loss, or major bugs.
98139
99-
cat review_response.json | jq -r ".choices[0].message.content" > review_output.txt
100-
'
140+
For each issue, use this exact format:
141+
```
142+
**[CRITICAL]** `File:Line` - Issue Title
143+
**Confidence:** High|Medium|Low (how sure you are this is a real problem)
144+
**Description:** Clear explanation of the issue
145+
**Impact:** What could go wrong if merged
146+
**Suggested Fix:** Specific code changes needed
147+
```
101148
102-
- name: Post review comment
103-
uses: actions/github-script@v7
104-
with:
105-
script: |
106-
const fs = require('fs');
107-
const reviewText = fs.readFileSync('review_output.txt', 'utf8');
108-
const formattedReview = `## OpenCode AI Review 🤖\n\n${reviewText}`;
149+
## ⚠️ High Priority Issues (Should Fix)
150+
Same approach as Critical - check previous reviews first, acknowledge fixes, only report unresolved issues.
151+
152+
Same format as Critical, but with **[HIGH]** prefix.
153+
154+
## 💡 Medium Priority Issues (Nice to Fix)
155+
Same approach - verify previous reports, acknowledge fixes, report only still-present issues.
156+
157+
Same format, with **[MEDIUM]** prefix.
158+
159+
## ℹ️ Low Priority Suggestions (Optional)
160+
Same approach.
109161
110-
await github.rest.issues.createComment({
111-
owner: context.repo.owner,
112-
repo: context.repo.repo,
113-
issue_number: context.issue.number,
114-
body: formattedReview
115-
});
162+
Same format, with **[LOW]** prefix.
163+
164+
## 📊 SOLID Principles Score
165+
| Principle | Score | Notes |
166+
|-----------|-------|-------|
167+
| Single Responsibility | 0-10 | Brief justification |
168+
| Open/Closed | 0-10 | Brief justification |
169+
| Liskov Substitution | 0-10 | Brief justification |
170+
| Interface Segregation | 0-10 | Brief justification |
171+
| Dependency Inversion | 0-10 | Brief justification |
172+
| **Average** | **X.X** | |
173+
174+
## 🎯 Final Assessment
175+
176+
### Overall Confidence Score: XX%
177+
Rate your confidence in this PR being ready to merge (0-100%).
178+
**How to interpret:**
179+
- 0-30%: Major concerns, do not merge without significant rework
180+
- 31-60%: Moderate concerns, several issues need addressing
181+
- 61-80%: Minor concerns, mostly ready with some fixes
182+
- 81-100%: High confidence, ready to merge or with trivial fixes
183+
184+
### Confidence Breakdown:
185+
- **Code Quality:** XX% (how well-written is the code?)
186+
- **Completeness:** XX% (does it fulfill requirements?)
187+
- **Risk Level:** XX% (how risky is this change?)
188+
- **Test Coverage:** XX% (are changes adequately tested?)
189+
190+
### Merge Readiness:
191+
- [ ] All critical issues resolved
192+
- [ ] SOLID average score >= 6.0
193+
- [ ] Overall confidence >= 60%
194+
- [ ] No security concerns
195+
- [ ] Tests present and passing (if applicable)
196+
197+
### Verdict:
198+
**MERGE** | **MERGE WITH FIXES** | **DO NOT MERGE**
199+
200+
One-sentence explanation of the verdict.
201+
202+
---
203+
204+
**Review Guidelines:**
205+
1. **MOST IMPORTANT:** Always check previous reviews and verify if issues are fixed before reporting them again
206+
2. Acknowledge fixes explicitly with ✅ **[FIXED]** markers
207+
3. Check the PR description for linked issues ("Fixes #123", "Closes #456", etc.) and verify the implementation
208+
4. Be extremely specific with file paths and line numbers
209+
5. Confidence scores should reflect how certain you are - use "Low" when unsure
210+
6. If you have nothing meaningful to add to a section, write "None identified" instead of omitting it
211+
7. Always provide actionable fixes, never just complaints

‎workflows/opencode/pr/opencode-pr.yml‎

Lines changed: 70 additions & 10 deletions
Original file line numberDiff line numberDiff line change
@@ -44,14 +44,70 @@ jobs:
4444
git config user.name "github-actions[bot]"
4545
git config user.email "github-actions[bot]@users.noreply.github.com"
4646
47+
- name: Fetch previous opencode reviews
48+
id: previous-reviews
49+
run: |
50+
# Get PR number
51+
PR_NUMBER="${{ github.event.pull_request.number }}"
52+
53+
# Fetch review comments from opencode-agent
54+
echo "Fetching previous automated reviews..."
55+
56+
# Get all reviews
57+
gh api /repos/${{ github.repository }}/pulls/${PR_NUMBER}/reviews \
58+
--jq '.[] | select(.user.login == "opencode-agent[bot]") | {body: .body, submitted_at: .submitted_at}' > /tmp/opencode_reviews.json 2>/dev/null || echo "[]" > /tmp/opencode_reviews.json
59+
60+
# Get all PR review comments (inline comments)
61+
gh api /repos/${{ github.repository }}/pulls/${PR_NUMBER}/comments \
62+
--jq '.[] | select(.user.login == "opencode-agent[bot]") | {body: .body, path: .path, line: .line, created_at: .created_at}' > /tmp/opencode_comments.json 2>/dev/null || echo "[]" > /tmp/opencode_comments.json
63+
64+
# Format the previous reviews for the prompt
65+
REVIEW_CONTENT="## Previous Automated Reviews from opencode-agent:\n\n"
66+
67+
# Process reviews
68+
if [ -s /tmp/opencode_reviews.json ] && [ "$(cat /tmp/opencode_reviews.json)" != "[]" ]; then
69+
while IFS= read -r review; do
70+
if [ -n "$review" ] && [ "$review" != "null" ]; then
71+
body=$(echo "$review" | jq -r '.body // empty')
72+
date=$(echo "$review" | jq -r '.submitted_at // empty')
73+
if [ -n "$body" ] && [ "$body" != "null" ]; then
74+
REVIEW_CONTENT="${REVIEW_CONTENT}### Review from ${date}\n${body}\n\n---\n\n"
75+
fi
76+
fi
77+
done < /tmp/opencode_reviews.json
78+
else
79+
REVIEW_CONTENT="${REVIEW_CONTENT}No previous automated reviews found.\n"
80+
fi
81+
82+
# Write to environment file using proper escaping
83+
echo "PREVIOUS_REVIEWS<<EOF" >> $GITHUB_ENV
84+
printf '%s\n' "$REVIEW_CONTENT" >> $GITHUB_ENV
85+
echo "EOF" >> $GITHUB_ENV
86+
87+
echo "Previous reviews fetched and formatted for context"
88+
env:
89+
GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
90+
4791
- name: Run opencode
4892
uses: anomalyco/opencode/github@latest
4993
env:
5094
KIMI_API_KEY: ${{ secrets.KIMI_API_KEY }}
5195
with:
5296
model: kimi-for-coding/k2p5
5397
prompt: |
54-
You are reviewing a pull request. Analyze the code changes and output your review in the following STRICT STRUCTURE:
98+
You are reviewing a pull request.
99+
100+
${{ env.PREVIOUS_REVIEWS }}
101+
102+
---
103+
104+
**YOUR TASK:** Analyze the CURRENT code changes and previous reviews above, then output your review in the following STRICT STRUCTURE:
105+
106+
**CRITICAL INSTRUCTIONS:**
107+
1. **CHECK PREVIOUS ISSUES FIRST:** Look at the "Previous Automated Reviews" section above. For each issue previously reported (Critical, High, Medium, Low), verify if it still exists in the current code.
108+
2. **ACKNOWLEDGE FIXES:** If a previously reported issue has been fixed, state "✅ **[FIXED]** Previous issue: [brief description]" in the appropriate section.
109+
3. **ONLY REPORT NEW/UNRESOLVED ISSUES:** Do NOT re-report issues that have already been fixed. Only report issues that are still present in the current code.
110+
4. **TRACK CHANGES:** If an issue was reported in a previous review but the code has changed, verify the new code and report the issue with updated file:line references if it still exists.
55111
56112
---
57113
@@ -66,7 +122,9 @@ jobs:
66122
Then provide 2-3 sentences summarizing the PR purpose, scope, and overall quality.
67123
68124
## 🔴 Critical Issues (Must Fix - Blocks Merge)
69-
Only issues that could cause crashes, security vulnerabilities, data loss, or major bugs.
125+
**IMPORTANT:** Check previous reviews first. If critical issues were reported before, verify if they're fixed. If fixed, say "✅ All previously reported critical issues have been resolved."
126+
127+
Only report NEW critical issues that could cause crashes, security vulnerabilities, data loss, or major bugs.
70128
71129
For each issue, use this exact format:
72130
```
@@ -78,17 +136,17 @@ jobs:
78136
```
79137
80138
## ⚠️ High Priority Issues (Should Fix)
81-
Significant code quality issues, potential bugs, or architectural problems.
139+
Same approach as Critical - check previous reviews first, acknowledge fixes, only report unresolved issues.
82140
83141
Same format as Critical, but with **[HIGH]** prefix.
84142
85143
## 💡 Medium Priority Issues (Nice to Fix)
86-
Style issues, minor optimizations, or code clarity improvements.
144+
Same approach - verify previous reports, acknowledge fixes, report only still-present issues.
87145
88146
Same format, with **[MEDIUM]** prefix.
89147
90148
## ℹ️ Low Priority Suggestions (Optional)
91-
Minor suggestions, documentation improvements, or subjective preferences.
149+
Same approach.
92150
93151
Same format, with **[LOW]** prefix.
94152
@@ -133,8 +191,10 @@ jobs:
133191
---
134192
135193
**Review Guidelines:**
136-
1. Check the PR description for linked issues ("Fixes #123", "Closes #456", etc.) and verify the implementation
137-
2. Be extremely specific with file paths and line numbers
138-
3. Confidence scores should reflect how certain you are - use "Low" when unsure
139-
4. If you have nothing meaningful to add to a section, write "None identified" instead of omitting it
140-
5. Always provide actionable fixes, never just complaints
194+
1. **MOST IMPORTANT:** Always check previous reviews and verify if issues are fixed before reporting them again
195+
2. Acknowledge fixes explicitly with ✅ **[FIXED]** markers
196+
3. Check the PR description for linked issues ("Fixes #123", "Closes #456", etc.) and verify the implementation
197+
4. Be extremely specific with file paths and line numbers
198+
5. Confidence scores should reflect how certain you are - use "Low" when unsure
199+
6. If you have nothing meaningful to add to a section, write "None identified" instead of omitting it
200+
7. Always provide actionable fixes, never just complaints

0 commit comments

Comments
 (0)