Skip to content

[DOCUMENTATION]: Reconcile SKILL.md with PR #112 changes (AETHERSCAN_EXTRA_BINDS, .env example) #137

Description

@claude

Context

PR #112 (merged as 13eb22c) added a new AETHERSCAN_EXTRA_BINDS env var to utils/run_container.sh, updated the README .env example block (reordered vars, added AETHERSCAN_EXTRA_BINDS), and updated the raw .h5 inference example to show the new var. KNOWN_ISSUES.md entry #16 was also added but neither CLAUDE.md nor the SKILL.md reference individual entries, so no drift there.

CLAUDE.md is unaffected — it doesn't restate any of the specific facts that changed.

.claude/skills/aetherscan-repo-context/SKILL.md has three drifts against the updated canonical docs (README.md):

Specific updates needed

SKILL.md edits (apply manually — .claude/ is read-only in CI)

Edit 1 — run_container.sh description (line 31): Add AETHERSCAN_EXTRA_BINDS to the description.

Old:

- **`utils/run_container.sh`** auto-detects apptainer vs singularity (Apptainer wins when both present), sets `--nv` for GPU passthrough, auto-loads `<repo>/.env`, and bind-mounts the repo + `AETHERSCAN_{DATA,MODEL,OUTPUT}_PATH` 1:1 so absolute paths persisted in the DB stay valid across host and container.

New:

- **`utils/run_container.sh`** auto-detects apptainer vs singularity (Apptainer wins when both present), sets `--nv` for GPU passthrough, auto-loads `<repo>/.env`, and bind-mounts the repo + `AETHERSCAN_{DATA,MODEL,OUTPUT}_PATH` 1:1 so absolute paths persisted in the DB stay valid across host and container. `AETHERSCAN_EXTRA_BINDS` (comma-separated host paths) appends additional 1:1 binds for data outside the standard dirs (e.g. raw `.h5` files under `/datag` for inference); the runtime's native `SINGULARITY_BIND` / `APPTAINER_BIND` still pass through and are additive.

Edit 2 — Raw .h5 inference example (lines 50-55): Add AETHERSCAN_EXTRA_BINDS prefix and update the comment to match README.md.

Old:

# Inference from raw .h5 (triggers energy-detection preprocessing)
./utils/run_container.sh python -m aetherscan.main inference \
    --inference-files complete_cadences_catalog.csv \
    --encoder-path /path/to/vae_encoder.keras \
    --rf-path /path/to/random_forest.joblib \
    --config-path /path/to/config.json

New:

# Inference from raw .h5 (triggers energy-detection preprocessing) — bind
# extra host paths if the .h5 files live outside the standard bind mounts
AETHERSCAN_EXTRA_BINDS=/datag ./utils/run_container.sh python -m aetherscan.main inference \
    --inference-files complete_cadences_catalog.csv \
    --encoder-path /path/to/vae_encoder.keras \
    --rf-path /path/to/random_forest.joblib \
    --config-path /path/to/config.json

Edit 3 — .env example block (lines 74-82): Add AETHERSCAN_EXTRA_BINDS and reorder to match README.md (path vars first, then extra binds, then Slack).

Old:

# .env example (Slack integration auto-disables if unset)
SLACK_BOT_TOKEN=your-slack-bot-token
SLACK_CHANNEL=your-slack-channel
# Defaults to /datax/scratch/zachy/{data|models|outputs}/aetherscan; CLI flags override
AETHERSCAN_DATA_PATH=/path/to/data
AETHERSCAN_MODEL_PATH=/path/to/models
AETHERSCAN_OUTPUT_PATH=/path/to/outputs

New:

# .env example
# Defaults to /datax/scratch/zachy/{data|models|outputs}/aetherscan; CLI flags override
AETHERSCAN_DATA_PATH=/path/to/data
AETHERSCAN_MODEL_PATH=/path/to/models
AETHERSCAN_OUTPUT_PATH=/path/to/outputs
# Optional: comma-separated extra host paths for run_container.sh to bind 1:1
AETHERSCAN_EXTRA_BINDS=/extra/host/paths
# Slack integration auto-disables if unset
SLACK_BOT_TOKEN=your-slack-bot-token
SLACK_CHANNEL=your-slack-channel

Instructions for @claude

Open a PR that Closes this issue. CLAUDE.md needs no changes. For SKILL.md:

  1. Do NOT attempt to modify .claude/skills/aetherscan-repo-context/SKILL.md — it is read-only in CI.
  2. Copy the three SKILL.md edits above verbatim into the PR body, followed by step-by-step instructions for a human to apply them: check out the PR branch locally (gh pr checkout <PR#>), make the listed edits to .claude/skills/aetherscan-repo-context/SKILL.md, commit, and push back to the branch for review & merge.
  3. Tag @zachtheyek (code owner per CODEOWNERS) in the PR body to perform the local apply.

Metadata

Metadata

Assignees

No one assigned

    Labels

    documentationImprovements or additions to documentation

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions