mirror of
https://github.com/allaunthefox/Research-Stack.git
synced 2026-07-31 03:05:21 +00:00
5 KiB
5 KiB
Evidence Attachment Guidelines
Systematic approach for attaching evidence to claims in the Research Stack.
Build Logs
Location: out/build_logs/
Naming Convention: lake_build_YYYYMMDD_UUID.log
Required Attachments:
- All build success claims must reference a build log file
- Build log must contain full lake build output with explicit step-by-step logging
- Each step must be logged with:
- UUID (unique identifier for the build step)
- Timestamp (ISO 8601 format)
- Step description
- Step output
- Build steps must be appended to a DAG structure
- Documentation must include markdown link to build log
DAG Structure:
{
"build_id": "UUID",
"timestamp": "YYYY-MM-DDTHH:MM:SSZ",
"commit": "full_commit_hash",
"steps": [
{
"step_id": "UUID",
"timestamp": "YYYY-MM-DDTHH:MM:SSZ",
"description": "step description",
"command": "command executed"
}
],
"final_timestamp": "YYYY-MM-DDTHH:MM:SSZ",
"status": "completed"
}
Implementation:
- Script:
scripts/dag_build_logger.sh - Generates UUID for each build and each step
- Logs each step with timestamp to both log file and DAG
- Saves DAG as JSON in
out/build_dag/ - Saves full log in
out/build_logs/
Path Quoting Invariant:
path_with_spaces → must_be_quoted
All filesystem paths must be quoted in shell scripts to handle spaces in repository paths. Use quoted variables:
REPO_ROOT="/home/allaun/Documents/Research Stack"
BUILD_LOG_DIR="$REPO_ROOT/out/build_logs"
mkdir -p "$BUILD_LOG_DIR"
Updated Files:
docs/specs/NII_CORE_DRIVER_COMPLETION_SUMMARY.md- referencesout/build_logs/lake_build_20260429.logdocs/nic_cpu_probe_report.md- referencesout/build_logs/lake_build_20260429.logdocs/FIELD_EQUATION_COMPARISON.md- referencesout/build_logs/lake_build_20260429.log
Benchmark Artifacts
Location: data/benchmarks/
Benchmark Artifacts:
braid_explained_events_1_200.csv(44,526 bytes) - Event sequence data for braid benchmarkbraid_feature_records_1_200.csv(23,662 bytes) - Feature extraction resultsbraid_field_interaction.png(95,097 bytes) - Field interaction visualizationbraid_glossary.csv(1,708 bytes) - Terminology referencebraid_photonic_emulation.csv(2,602 bytes) - Photonic emulation databraid_photonic_emulation.png(79,633 bytes) - Photonic emulation visualizationbraid_photonic_netlist.json(12,545 bytes) - Netlist representationkiller_criterion/(directory with FASTA files for killer criterion benchmark)
Required Attachments:
- All benchmark performance claims must reference specific artifact files
- Artifacts must include: data files, methodology documentation, environment configuration
- Documentation must include markdown links to artifacts
Files Requiring Updates:
0-Core-Formalism/lean/Semantics/ChatGPT-Hutter_prize_Compression_#1.md- references braid benchmark artifacts- Any documentation referencing benchmark results must link to specific artifact files
Citation Verification
Required:
- All citation claims must include accessible source URLs
- Verify URLs are stable and retrievable
- Document citation context and relevance
- Ensure sources are not ephemeral (avoid link rot)
Verification Process:
- Extract all citation URLs from documentation
- Test URL accessibility with curl/wget
- Document any broken or unstable links
- Replace unstable links with stable alternatives or archived versions
Commit References
Current Commit: ddf028ab (HEAD -> reorg/lean-domain-atlas-2026-04-29)
Required:
- All commit references must use full commit hashes (40 characters)
- Document commit context and message
- Verify commit exists in repository
- Avoid branch names or partial hashes
Automated Evidence Attachment
Implementation Plan:
- Create pre-commit hook to generate build logs
- Create script to verify citation accessibility
- Create script to validate commit references
- Integrate evidence attachment into CI/CD pipeline
Invariant Enforcement:
buildClaim → buildLog
benchmarkClaim → benchmarkArtifact
citationClaim → sourceCitation
commitClaim → commitHash
User Bandwidth Protection
Core Law: If a verification claim consumes user attention to validate the claim itself, the workflow has already failed.
Implementation:
- All claims must have attached evidence at claim time
- User should never need to research to validate a claim
- Evidence must be in the repository or reliably accessible
- No "trust me, I verified this" claims without artifacts