Oracle

oracle — best use

Oracle bundles your prompt + selected files into one “one-shot” request so another model can answer with real repo context (API or browser automation). Treat output as advisory: verify against code + tests.

Main use case (browser, GPT‑5.2 Pro)

Default workflow here: --engine browser with GPT‑5.2 Pro in ChatGPT. This is the common “long think” path: ~10 minutes to ~1 hour is normal; expect a stored session you can reattach to.

Recommended defaults:

• Engine: browser (--engine browser)
• Model: GPT‑5.2 Pro (--model gpt-5.2-pro or --model "5.2 Pro")

Golden path

1. Pick a tight file set (fewest files that still contain the truth). 2. Preview payload + token spend (--dry-run + --files-report). 3. Use browser mode for the usual GPT‑5.2 Pro workflow; use API only when you explicitly want it. 4. If the run detaches/timeouts: reattach to the stored session (don’t re-run).

Commands (preferred)

• Help:
• oracle --help
• If the binary isn’t installed: npx -y @steipete/oracle --help (avoid pnpx here; sqlite bindings).

• Preview (no tokens):
• oracle --dry-run summary -p "<task>" --file "src/**" --file "!**/*.test.*"
• oracle --dry-run full -p "<task>" --file "src/**"

• Token sanity:
• oracle --dry-run summary --files-report -p "<task>" --file "src/**"

• Browser run (main path; long-running is normal):
• oracle --engine browser --model gpt-5.2-pro -p "<task>" --file "src/**"

• Manual paste fallback:
• oracle --render --copy -p "<task>" --file "src/**"
• Note: --copy is a hidden alias for --copy-markdown.

Attaching files (--file)

--file accepts files, directories, and globs. You can pass it multiple times; entries can be comma-separated.

• Include:
• --file "src/**"
• --file src/index.ts
• --file docs --file README.md

• Exclude:
• --file "src/**" --file "!src/**/*.test.ts" --file "!**/*.snap"

• Defaults (implementation behavior):
• Default-ignored dirs: node_modules, dist, coverage, .git, .turbo, .next, build, tmp (skipped unless explicitly passed as literal dirs/files).
• Honors .gitignore when expanding globs.
• Does not follow symlinks.
• Dotfiles filtered unless opted in via pattern (e.g. --file ".github/**").
• Files > 1 MB rejected.

Engines (API vs browser)

• Auto-pick: api when OPENAI_API_KEY is set; otherwise browser.
• Browser supports GPT + Gemini only; use --engine api for Claude/Grok/Codex or multi-model runs.
• Browser attachments:
• --browser-attachments auto|never|always (auto pastes inline up to ~60k chars then uploads).
• Remote browser host:
• Host: oracle serve --host 0.0.0.0 --port 9473 --token <secret>
• Client: oracle --engine browser --remote-host <host:port> --remote-token <secret> -p "<task>" --file "src/**"

Sessions + slugs

• Stored under ~/.oracle/sessions (override with ORACLE_HOME_DIR).
• Runs may detach or take a long time (browser + GPT‑5.2 Pro often does). If the CLI times out: don’t re-run; reattach.
• List: oracle status --hours 72
• Attach: oracle session <id> --render
• Use --slug "<3-5 words>" to keep session IDs readable.
• Duplicate prompt guard exists; use --force only when you truly want a fresh run.

Prompt template (high signal)

Oracle starts with zero project knowledge. Assume the model cannot infer your stack, build tooling, conventions, or “obvious” paths. Include:

• Project briefing (stack + build/test commands + platform constraints).
• “Where things live” (key directories, entrypoints, config files, boundaries).
• Exact question + what you tried + the error text (verbatim).
• Constraints (“don’t change X”, “must keep public API”, etc).
• Desired output (“return patch plan + tests”, “give 3 options with tradeoffs”).

Safety

• Don’t attach secrets by default (.env, key files, auth tokens). Redact aggressively; share only what’s required.

“Exhaustive prompt” restoration pattern

For long investigations, write a standalone prompt + file set so you can rerun days later:

• 6–30 sentence project briefing + the goal.
• Repro steps + exact errors + what you tried.
• Attach all context files needed (entrypoints, configs, key modules, docs).

Oracle runs are one-shot; the model doesn’t remember prior runs. “Restoring context” means re-running with the same prompt + --file … set (or reattaching a still-running stored session).