fix(memory): persist verified host toolchain across restarts - #323
Closed
parkjs101 wants to merge 4 commits into
Closed
fix(memory): persist verified host toolchain across restarts#323parkjs101 wants to merge 4 commits into
parkjs101 wants to merge 4 commits into
Conversation
|
Important Review skippedDraft detected. Please check the settings in the CodeRabbit UI or the ⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Plus Run ID: You can disable this status message by setting the Use the checkbox below for a quick retry:
📝 WalkthroughWalkthroughChangesDurable host toolchain
Estimated code review effort: 4 (Complex) | ~45 minutes Sequence Diagram(s)sequenceDiagram
participant MemoryRuntime
participant HostToolchain
participant ProfileFile
participant MemoryIndex
participant DiskPromptBuilder
MemoryRuntime->>HostToolchain: resolve and refresh profile
HostToolchain->>ProfileFile: merge managed JSON block
MemoryRuntime->>MemoryIndex: reindex profile.md
DiskPromptBuilder->>ProfileFile: read and parse profile
DiskPromptBuilder->>HostToolchain: render bounded prompt block
Suggested reviewers: 🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✨ Finishing Touches 💡 1⚔️ Resolve merge conflicts 💡
📝 Generate docstrings
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
…oolchain # Conflicts: # src/memory/host-toolchain.ts # src/prompt/builder.ts # structure/str_func.md
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 4
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@src/memory/host-toolchain.ts`:
- Line 456: Update the truncation logic in the host-toolchain rendering flow to
ensure the returned string never exceeds maxChars, including when maxChars is
smaller than the truncation suffix length. Preserve the existing full-rendered
result when it fits, and shorten or omit the suffix as needed for small budgets.
- Around line 219-226: Limit PATH-derived candidates per tool before
resolveHostToolchainProfile verifies them, using a small fixed cap applied to
the results from listCliBinaryCandidates. Preserve scanError handling and
candidate deduplication, and ensure excess candidates are excluded from
verification.
In `@src/prompt/builder.ts`:
- Around line 731-732: Update the AGENTS.md generation flow around
loadDiskHostToolchain() to refresh or regenerate the host-toolchain profile
before loading it, ensuring profile.md contains current verified absolute paths.
Preserve the existing conditional prompt append behavior after the refreshed
profile is loaded.
In `@structure/str_func.md`:
- Around line 210-222: Synchronize the documented src/memory inventory: in
structure/str_func.md lines 210-222, change the module count from 15 to 16 and
add shared.ts, synonyms.ts, worklog.ts, and heartbeat-report.ts; in
structure/memory_architecture.md lines 11-12, add src/memory/heartbeat-report.ts
to the source inventory.
🪄 Autofix
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro Plus
Run ID: 2a640be7-d9b7-4693-8555-7393eafd97cf
📒 Files selected for processing (8)
src/memory/bootstrap.tssrc/memory/host-toolchain.tssrc/memory/runtime.tssrc/prompt/builder.tsstructure/infra.mdstructure/memory_architecture.mdstructure/str_func.mdtests/unit/host-toolchain-profile.test.ts
Comment on lines
+219
to
+226
| for (const envName of ENV_PATHS[tool]) addCandidate(candidates, env[envName], `env:${envName}`); | ||
|
|
||
| let scanError = false; | ||
| for (const name of pathNames(tool, platform)) { | ||
| const scan = listCliBinaryCandidates(name, env['PATH'] || env['Path'] || env['path'] || ''); | ||
| scanError ||= !!scan.scanError; | ||
| for (const candidate of scan.candidates) addCandidate(candidates, candidate.path, 'PATH'); | ||
| } |
There was a problem hiding this comment.
🚀 Performance & Scalability | 🟡 Minor | ⚡ Quick win
🧩 Analysis chain
🏁 Script executed:
#!/usr/bin/env bash
set -euo pipefail
fd -a -t f 'cli-detect\.(ts|mts|cts|js)$' src/core | while IFS= read -r file; do
echo "== $file =="
ast-grep outline "$file" --items all
rg -n -C 10 'listCliBinaryCandidates|scanError|candidates|slice\(|limit' "$file"
doneRepository: lidge-jun/cli-jaw
Length of output: 12663
🏁 Script executed:
#!/usr/bin/env bash
set -euo pipefail
echo '== host-toolchain symbols =='
ast-grep outline src/memory/host-toolchain.ts --items all
rg -n -C 12 'listCliBinaryCandidates|spawnSync|1[.]8|timeout|resolveHostToolchainProfile|scanError' src/memory/host-toolchain.tsRepository: lidge-jun/cli-jaw
Length of output: 9718
Enforce a per-tool PATH candidate limit.
listCliBinaryCandidates returns every unique which -a result. resolveHostToolchainProfile verifies each candidate sequentially, with an 1,800 ms timeout per --version probe. Cap the candidates before verification, or add an equivalent limit in listCliBinaryCandidates.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@src/memory/host-toolchain.ts` around lines 219 - 226, Limit PATH-derived
candidates per tool before resolveHostToolchainProfile verifies them, using a
small fixed cap applied to the results from listCliBinaryCandidates. Preserve
scanError handling and candidate deduplication, and ensure excess candidates are
excluded from verification.
| if (record?.lastAttemptAt) lines.push('', `verified_at: ${record.lastAttemptAt}`); | ||
| return lines.join('\n'); | ||
| const rendered = lines.join('\n'); | ||
| return rendered.length <= maxChars ? rendered : `${rendered.slice(0, Math.max(0, maxChars - 16))}\n...(truncated)`; |
There was a problem hiding this comment.
🎯 Functional Correctness | 🟡 Minor | ⚡ Quick win
Honor maxChars for small budgets.
Line 456 returns the full truncation suffix when maxChars is less than 16. The result can exceed the caller budget.
Proposed fix
- const rendered = lines.join('\n');
- return rendered.length <= maxChars ? rendered : `${rendered.slice(0, Math.max(0, maxChars - 16))}\n...(truncated)`;
+ const rendered = lines.join('\n');
+ const budget = Math.max(0, maxChars);
+ const suffix = '\n...(truncated)';
+ if (rendered.length <= budget) return rendered;
+ if (budget <= suffix.length) return suffix.slice(0, budget);
+ return `${rendered.slice(0, budget - suffix.length)}${suffix}`;📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
Suggested change
| return rendered.length <= maxChars ? rendered : `${rendered.slice(0, Math.max(0, maxChars - 16))}\n...(truncated)`; | |
| const rendered = lines.join('\n'); | |
| const budget = Math.max(0, maxChars); | |
| const suffix = '\n...(truncated)'; | |
| if (rendered.length <= budget) return rendered; | |
| if (budget <= suffix.length) return suffix.slice(0, budget); | |
| return `${rendered.slice(0, budget - suffix.length)}${suffix}`; |
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@src/memory/host-toolchain.ts` at line 456, Update the truncation logic in the
host-toolchain rendering flow to ensure the returned string never exceeds
maxChars, including when maxChars is smaller than the truncation suffix length.
Preserve the existing full-rendered result when it fits, and shorten or omit the
suffix as needed for small budgets.
Comment on lines
+731
to
+732
| const hostToolchain = loadDiskHostToolchain(); | ||
| if (hostToolchain) prompt += `\n\n---\n${hostToolchain}\n`; |
There was a problem hiding this comment.
🗄️ Data Integrity & Integration | 🟡 Minor | ⚡ Quick win
🧩 Analysis chain
🏁 Script executed:
#!/usr/bin/env bash
set -euo pipefail
rg -n -C 12 -i \
'getSystemPrompt|host.?toolchain|profile\.md|bootstrap|refresh' \
src/memory src/prompt testsRepository: lidge-jun/cli-jaw
Length of output: 50374
🏁 Script executed:
#!/usr/bin/env bash
set -euo pipefail
printf '%s\n' '== relevant symbols =='
rg -n \
'function getSystemPrompt|const getSystemPrompt|export .*getSystemPrompt|loadDiskHostToolchain|resolveHostToolchainProfile|scanSystemProfile|profile\.md|bootstrap' \
src/prompt/builder.ts src/memory src/agent src/core tests/unit/host-toolchain-profile.test.ts
printf '%s\n' '== builder structure =='
ast-grep outline src/prompt/builder.ts --match 'getSystemPrompt' --view expanded
printf '%s\n' '== host-toolchain module structure =='
ast-grep outline src/memory/host-toolchain.ts --view expandedRepository: lidge-jun/cli-jaw
Length of output: 13015
🏁 Script executed:
#!/usr/bin/env bash
set -euo pipefail
printf '%s\n' '== prompt loader and renderer =='
sed -n '640,745p' src/prompt/builder.ts
printf '%s\n' '== host-toolchain parsing, refresh, and rendering =='
sed -n '250,455p' src/memory/host-toolchain.ts
printf '%s\n' '== memory initialization and bootstrap ordering =='
sed -n '275,355p' src/memory/runtime.ts
sed -n '380,435p' src/memory/bootstrap.ts
sed -n '545,600p' src/memory/bootstrap.ts
printf '%s\n' '== all getSystemPrompt call sites =='
rg -n -C 8 'getSystemPrompt\s*\(' --glob '!src/prompt/builder.ts' --glob '!tests/**' .Repository: lidge-jun/cli-jaw
Length of output: 50375
🏁 Script executed:
#!/usr/bin/env bash
set -euo pipefail
printf '%s\n' '== disk prompt generation =='
rg -n -C 20 'function regenerateB|export .*regenerateB|getSystemPrompt\(\{ forDisk: true \}\)|getSystemPrompt\(\{[^}]*forDisk' src tests --glob '*.ts'
printf '%s\n' '== memory readiness callers =='
rg -n -C 12 'ensureIntegratedMemoryReady\(|refreshHostToolchain\(|refreshHostToolchainProfileFile\(' src tests --glob '*.ts'
printf '%s\n' '== startup and command orchestration =='
rg -n -C 16 'regenerateB\(|ensureIntegratedMemoryReady\(|initPromptFiles\(' src/cli src/core src/agent src/index.ts src/main.ts bin tests --glob '*.ts' 2>/dev/null || trueRepository: lidge-jun/cli-jaw
Length of output: 50373
🏁 Script executed:
#!/usr/bin/env bash
set -euo pipefail
printf '%s\n' '== ensureIntegratedMemoryReady references =='
rg -n 'ensureIntegratedMemoryReady' . --glob '*.ts' --glob '!tests/**'
printf '%s\n' '== regenerateB references =='
rg -n 'regenerateB' src bin --glob '*.ts'
printf '%s\n' '== runtime settings refresh region =='
sed -n '330,410p' src/core/runtime-settings.ts
printf '%s\n' '== entrypoint candidates and memory startup calls =='
git ls-files | rg '(^|/)(main|index|startup|bootstrap|runtime|app)[^/]*\.(ts|js)$' | head -100
rg -n -C 10 'ensureIntegratedMemoryReady|regenerateB|initPromptFiles' src bin --glob '*.ts' --glob '!src/prompt/builder.ts'Repository: lidge-jun/cli-jaw
Length of output: 42558
🏁 Script executed:
#!/usr/bin/env bash
set -euo pipefail
python3 - <<'PY'
from pathlib import Path
import re
builder = Path("src/prompt/builder.ts").read_text()
runtime = Path("src/memory/runtime.ts").read_text()
spawn = Path("src/agent/spawn.ts").read_text()
runtime_settings = Path("src/core/runtime-settings.ts").read_text()
get_prompt = re.search(
r"export function getSystemPrompt\b[\s\S]*?(?=\nexport function |\nfunction |\nlet _lastPromptHash)",
builder,
)
regenerate = re.search(
r"export function regenerateB\b[\s\S]*?(?=\nexport function |\nfunction |\nlet )",
builder,
)
assert get_prompt and regenerate
prompt_body = get_prompt.group(0)
regenerate_body = regenerate.group(0)
print("getSystemPrompt reads profile.md:", "readFileSync(profilePath" in prompt_body)
print("getSystemPrompt refreshes profile:", "refreshHostToolchainProfileFile" in prompt_body or "ensureIntegratedMemoryReady" in prompt_body)
print("regenerateB calls disk prompt:", "getSystemPrompt({ forDisk: true })" in regenerate_body)
print("spawn refreshes via regenerateB:", "regenerateB();" in spawn)
print("runtime-settings refreshes before regenerateB:",
bool(re.search(r"refreshHostToolchainProfileFile[\s\S]{0,500}regenerateB\(\)", runtime_settings)))
print("ensureIntegratedMemoryReady direct references:",
len(re.findall(r"\bensureIntegratedMemoryReady\b", runtime)))
PYRepository: lidge-jun/cli-jaw
Length of output: 244
🏁 Script executed:
#!/usr/bin/env bash
set -euo pipefail
python3 - <<'PY'
from pathlib import Path
builder = Path("src/prompt/builder.ts").read_text()
runtime = Path("src/memory/runtime.ts").read_text()
spawn = Path("src/agent/spawn.ts").read_text()
runtime_settings = Path("src/core/runtime-settings.ts").read_text()
def section(text, start_marker, end_marker=None):
start = text.index(start_marker)
end = text.index(end_marker, start) if end_marker else len(text)
return text[start:end]
prompt_body = section(builder, "export function getSystemPrompt", "export function getEmployeePrompt")
regenerate_body = section(builder, "export function regenerateB")
checks = {
"getSystemPrompt reads profile.md": "readFileSync(profilePath" in prompt_body,
"getSystemPrompt refreshes profile": (
"refreshHostToolchainProfileFile" in prompt_body
or "ensureIntegratedMemoryReady" in prompt_body
),
"regenerateB calls disk prompt": "getSystemPrompt({ forDisk: true })" in regenerate_body,
"spawn refreshes via regenerateB": "regenerateB();" in spawn,
"runtime-settings calls regenerateB": "regenerateB();" in runtime_settings,
"runtime-settings calls host-toolchain refresh": (
"refreshHostToolchainProfileFile" in runtime_settings
or "ensureIntegratedMemoryReady" in runtime_settings
),
}
for name, value in checks.items():
print(f"{name}: {value}")
print(
"ensureIntegratedMemoryReady references:",
runtime.count("ensureIntegratedMemoryReady"),
)
assert checks["getSystemPrompt reads profile.md"]
assert not checks["getSystemPrompt refreshes profile"]
assert checks["regenerateB calls disk prompt"]
assert checks["spawn refreshes via regenerateB"]
assert checks["runtime-settings calls regenerateB"]
assert not checks["runtime-settings calls host-toolchain refresh"]
PYRepository: lidge-jun/cli-jaw
Length of output: 535
🏁 Script executed:
#!/usr/bin/env bash
set -euo pipefail
python3 - <<'PY'
from pathlib import Path
builder = Path("src/prompt/builder.ts").read_text()
runtime = Path("src/memory/runtime.ts").read_text()
spawn = Path("src/agent/spawn.ts").read_text()
runtime_settings = Path("src/core/runtime-settings.ts").read_text()
disk_loader = builder[builder.index("function loadDiskHostToolchain"):
builder.index("function getCurrentSessionIdentityLine")]
regenerate = builder[builder.index("export function regenerateB"):]
checks = {
"loadDiskHostToolchain reads profile.md": "readFileSync(profilePath" in disk_loader,
"loadDiskHostToolchain refreshes profile": (
"refreshHostToolchainProfileFile" in disk_loader
or "ensureIntegratedMemoryReady" in disk_loader
),
"regenerateB calls disk prompt": "getSystemPrompt({ forDisk: true })" in regenerate,
"spawn calls regenerateB": "regenerateB();" in spawn,
"runtime-settings calls regenerateB": "regenerateB();" in runtime_settings,
"runtime-settings calls host-toolchain refresh": (
"refreshHostToolchainProfileFile" in runtime_settings
or "ensureIntegratedMemoryReady" in runtime_settings
),
}
for name, value in checks.items():
print(f"{name}: {value}")
print("ensureIntegratedMemoryReady references:", runtime.count("ensureIntegratedMemoryReady"))
print("refreshHostToolchainProfileFile references:", runtime.count("refreshHostToolchainProfileFile"))
assert checks["loadDiskHostToolchain reads profile.md"]
assert not checks["loadDiskHostToolchain refreshes profile"]
assert checks["regenerateB calls disk prompt"]
assert checks["spawn calls regenerateB"]
assert checks["runtime-settings calls regenerateB"]
assert not checks["runtime-settings calls host-toolchain refresh"]
assert runtime.count("ensureIntegratedMemoryReady") == 2
PYRepository: lidge-jun/cli-jaw
Length of output: 495
Refresh the host-toolchain profile before regenerating AGENTS.md. loadDiskHostToolchain() only reads profile.md, so stale verified absolute paths can be written to AGENTS.md.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@src/prompt/builder.ts` around lines 731 - 732, Update the AGENTS.md
generation flow around loadDiskHostToolchain() to refresh or regenerate the
host-toolchain profile before loading it, ensuring profile.md contains current
verified absolute paths. Preserve the existing conditional prompt append
behavior after the refreshed profile is loaded.
Comment on lines
+210
to
+222
| │ ├── memory/ ← 데이터 영속화 + advanced memory runtime (15 files) | ||
| │ │ ├── advanced.ts ← Advanced Memory re-export stub (1L) | ||
| │ │ ├── bootstrap.ts ← legacy memory/bootstrap import + structured root 초기화 (584L) | ||
| │ │ ├── bootstrap.ts ← legacy memory/bootstrap import + structured root 초기화 (597L) | ||
| │ │ ├── heartbeat.ts ← Heartbeat 잡 스케줄 + cron/every timer orchestration + minute-slot dedupe + fs.watch (311L) | ||
| │ │ ├── heartbeat-schedule.ts ← Heartbeat schedule normalize + cron validate/match + timezone validate + immediate cron loop helper (410L) | ||
| │ │ ├── host-toolchain.ts ← durable verified host paths + profile managed block + bounded AGENTS summary (457L) | ||
| │ │ ├── identity.ts ← `shared/soul.md` 관리 + soul runtime helper (87L) | ||
| │ │ ├── indexing.ts ← FTS5/BM25 reindex + indexed file/chunk 상태 집계 (721L) | ||
| │ │ ├── injection.ts ← memory injection policy + advanced/basic search routing (69L) | ||
| │ │ ├── keyword-expand.ts ← search keyword expansion + provider config normalize (98L) | ||
| │ │ ├── memory.ts ← Persistent Memory grep 기반 (165L) | ||
| │ │ ├── reflect.ts ← episode → shared/procedures reflection + promoted fact 정리 (380L) | ||
| │ │ ├── runtime.ts ← Advanced Memory 런타임: bootstrap/import/FTS5 인덱스/BM25 검색/task snapshot/delta reindex (380L) | ||
| │ │ ├── runtime.ts ← Advanced Memory 런타임: bootstrap/import/FTS5 인덱스/BM25 검색/task snapshot/delta reindex (396L) |
There was a problem hiding this comment.
📐 Maintainability & Code Quality | 🟡 Minor | ⚡ Quick win
Synchronize the src/memory inventory.
The documentation has inconsistent module counts and file lists.
structure/str_func.md#L210-L222: Change the count to 16. Addshared.ts,synonyms.ts,worklog.ts, andheartbeat-report.ts.structure/memory_architecture.md#L11-L12: Addsrc/memory/heartbeat-report.tsto the source inventory.
📍 Affects 2 files
structure/str_func.md#L210-L222(this comment)structure/memory_architecture.md#L11-L12
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@structure/str_func.md` around lines 210 - 222, Synchronize the documented
src/memory inventory: in structure/str_func.md lines 210-222, change the module
count from 15 to 16 and add shared.ts, synonyms.ts, worklog.ts, and
heartbeat-report.ts; in structure/memory_architecture.md lines 11-12, add
src/memory/heartbeat-report.ts to the source inventory.
Collaborator
Author
|
Closing without merge. Issue #299 is already completed on dev by ca2426f with accepted CI-backed behavior. This PR replaces that accepted sidecar implementation with a larger profile.md redesign, so it is no longer required for the resolved issue. Keeping the branch available for reference; no branch deletion. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
officecli,soffice, Python, and ripgrep## Host toolchainblock in generated diskAGENTS.mdRoot cause
The system scan recorded general runtime facts but no durable document/search tool paths. Generated disk prompts therefore gave rebooted agents no authoritative host-toolchain state, forcing discovery from zero on every session and allowing one failed probe to be misreported as an absent installation.
Design
The durable record lives inside
memory/structured/profile.mdbetween cli-jaw-owned markers. It stores only absolute paths, safe version tokens, source labels,verified_at, and a verification result. Startup verifies a cached path directly before any discovery. The raw managed JSON is removed from normal profile summaries; disk prompt generation renders a separate block capped at 1800 characters.Current
devgained a smaller sidecar-based implementation while this branch was in progress. This branch is merged with currentdevand replaces that sidecar/duplicate startup probe with the profile-backed contract above.This intentionally does not add a generic shell runner or broaden Windows shell-resolution policy (#302/#310).
Addresses #299 without closing it.
Validation
npx tsx --experimental-test-module-mocks --test tests/unit/host-toolchain-profile.test.ts tests/unit/memory-core-profile-sync.test.ts tests/unit/agents-md-soul.test.ts— 12 passednpm run typecheck— passed after merging currentdevnpm run build— passed after merging currentdev; atomicdist/swap and asset verification succeededdist/contains the host-toolchain resolver/refresh symbolsbash structure/verify-counts.shwith a clean WSL systemPATH— 418 checks passednpm run test:all— attempted, but this host has unrelated existing browser/clone failures and the run blocks indefinitely on the interactive WindowsFolderBrowserDialogworkspace-picker test because the runner uses--test-timeout=0; the exact spawned test process tree was stopped. No full-suite pass is claimed.