engram
Design Mockup 中文 ← All pages

Context Preview

Simulate exactly what the LLM will see before any LLM is invoked. The primary diagnostic page for the Relevance Gate.

Task description

Token budget 4,000 tokens
50032 000
Include scopes
LAST RUN — 2.3 s
Candidates scored
62
Included
14
Excluded
48
Tokens used
3,847

Packed prompt preview

3,847 / 4,000 tokens
# ENGRAM CONTEXT PACK # Task: Review and merge PR #448: add retry logic to payment provider calls in acme-checkout-service # Model: Claude Sonnet 4 · Budget: 4000 tokens · Packed: 3847 tokens # Scopes: project, user, team, pool # Generated: 2026-04-18T14:32:01Z # Assets included: 14 of 62 candidates --- ## [MEMORY:project/feedback_push_confirm] [score:0.87] [tokens:210] id: feedback_push_confirm subtype: user, enforcement: default Always ask for explicit confirmation before running `git push` or any command that writes to a remote. This preference applies to all branches. --- ## [MEMORY:project/payment_provider_retry_policy] [score:0.96] [tokens:340] id: payment_provider_retry_policy subtype: feedback, enforcement: mandatory Payment provider calls must use exponential backoff with jitter. Max retries: 3. Base delay: 200ms. Max delay: 5s. Idempotency keys required for all POST requests. --- ## [MEMORY:project/pr_review_checklist] [score:0.91] [tokens:285] id: pr_review_checklist subtype: feedback, enforcement: default Code review checklist: 1. Tests added for new code paths 2. Error handling covers timeout and 5xx responses 3. No hardcoded credentials or URLs 4. Circuit breaker configured if calling external APIs 5. Metrics instrumentation for new service calls --- ## [MEMORY:team/api_versioning_policy] [score:0.79] [tokens:198] id: api_versioning_policy subtype: feedback, enforcement: mandatory, scope: team All external API clients must pin to a specific API version. Breaking changes to the payment provider API require a separate PR. --- ## [MEMORY:project/sprint_context_2026q2] [score:0.74] [tokens:420] id: sprint_context_2026q2 subtype: project, enforcement: hint Sprint goal Q2 2026: Improve payment reliability to 99.95% success rate. Current: 99.87%. Root cause: transient network errors not retried. PR #448 is the primary deliverable for this sprint. --- ## [WORKFLOW_PTR:release-checklist] [score:0.71] [tokens:95] ... # [8 more assets follow — 2,299 tokens remaining shown below] # BOUNDARY ↑ included / excluded ↓ (budget exhausted)

Relevance Gate — candidate ranking

62 scored · cutoff at rank 14
# Asset ID Subtype Scope Score Tokens Cumul. Inclusion reason
1 payment_provider_retry_policy feedback project 0.96 340 340 keyword match: "payment provider", "retry"
2 pr_review_checklist feedback project 0.91 285 625 keyword match: "PR", "review"; mandatory tag
3 feedback_push_confirm user project 0.87 210 835 high usage_count (34); always included for git ops
4 circuit_breaker_config feedback project 0.85 188 1,023 keyword match: "circuit breaker"; related to retry
5 checkout_service_sla reference project 0.82 165 1,188 inbound ref from payment_provider_retry_policy
6 api_versioning_policy feedback team 0.79 198 1,386 mandatory team rule; external API involved
7 sprint_context_2026q2 project project 0.74 420 1,806 keyword match: "payment reliability", "PR #448"
8 release-checklist (ptr) workflow_ptr project 0.71 95 1,901 workflow_ptr always included when scope=project
9 idempotency_key_spec reference pool 0.68 240 2,141 keyword match: "idempotency"; from compliance-checklists pool
10 error_handling_guide feedback team 0.65 312 2,453 semantic similarity: error handling patterns
11 payment_provider_docs_ptr reference project 0.62 88 2,541 inbound ref from payment_provider_retry_policy
12 testing_conventions feedback project 0.60 265 2,806 PR review checklist co-loaded 88% of sessions
13 observability_standards feedback team 0.58 580 3,386 metrics mentioned in checklist; fits remaining budget
14 user_profile_sky user user 0.55 461 3,847 user profile always in tail if budget allows
↑ Included (3,847 tokens)  ·  Budget exhausted — 48 candidates below this line were scored but excluded
15 staging_environment_ref reference project 0.52 320 budget exhausted
16 old_deploy_runbook project project 0.48 890 budget exhausted
17 frontend_styling_guide feedback team 0.11 210 score below 0.15 threshold; unrelated to task
Relevance Gate run in 2.3 s · 62 assets scored · debounced 500 ms per keystroke