Skip to content

WORKFLOWS.md

Latest commit

 

History

History
137 lines (96 loc) · 3.86 KB

WORKFLOWS.md

File metadata and controls

137 lines (96 loc) · 3.86 KB
Raw

Novel OS — Workflows

Step-by-step recipes for using Novel OS day to day.

🚀 First-time setup

git clone https://github.com/mrigankad/Novel-OS.git
cd Novel-OS
pip install -r requirements.txt
cp .env.example .env
# edit .env and set ONE provider key (Anthropic / OpenAI / Gemini / Kimi / NVIDIA / etc.)

The first key found is auto-detected. Override with NOVEL_OS_LLM_PROVIDER if you have several keys set.

📚 Standard chapter loop

flowchart LR
    P[plan chapter] --> W[write]
    W --> E[edit]
    E --> Ck[check]
    Ck --> V[validate]
    V --> A[approve]
    A -.next.-> P
Loading 
# 1. Plan — Architect expands the chapter outline
python core/orchestrator.py plan chapter --number 1 --pov "Lena Vasquez"

# 2. Write — Scribe drafts the prose
python core/orchestrator.py write --chapter 1

# 3. Edit — Editor refines (pick a mode)
python core/orchestrator.py edit --chapter 1 --mode line
# Modes: line | developmental | pacing | dialogue | tension

# 4. Check — free deterministic continuity scan
python core/orchestrator.py check --chapter 1

# 5. Validate — LLM Guardian validates (pre-check findings are included automatically)
python core/orchestrator.py validate --chapter 1

# 6. Approve — gated on Status: FAIL
python core/orchestrator.py approve --chapter 1

🧪 Dry-run mode (no API calls)

Every LLM-backed command supports --dry-run. It writes the prompt to disk so you can run it yourself in a chat UI, then submit the output via the --draft-file / --edited-file flags.

python core/orchestrator.py write --chapter 1 --dry-run
# -> outputs/chapter_001_scribe_prompt.md

# Run that prompt in ChatGPT/Claude/etc, save the response, then:
python core/orchestrator.py write --chapter 1 --draft-file my_response.md

Submitted files still get parsed — [SCRIBE_STATE_UPDATE] / [EDITOR_STATE_UPDATE] blocks update state regardless of how the chapter was produced.

🔬 Auditing an existing project

python core/orchestrator.py status            # dashboard
python core/orchestrator.py check             # whole-project continuity scan
python core/orchestrator.py check --chapter 12   # as-of a specific chapter
python core/orchestrator.py character list    # full cast with arc state
python core/orchestrator.py plot list         # threads by priority

The continuity engine exits non-zero on critical findings — wire it into CI if you care.

✍️ Hand-edited prose

Want to edit a chapter manually instead of running the Editor agent?

# Skip 'edit' entirely, then:
python core/orchestrator.py write --chapter N --draft-file my_revised.md
# Or replace the file directly:
# outputs/manuscript/chapter_NNN_revised.md
python core/orchestrator.py validate --chapter N   # Guardian still runs
python core/orchestrator.py approve  --chapter N

🔄 Re-running a phase

Phases are idempotent — re-running overwrites the corresponding artifact and re-parses.

python core/orchestrator.py edit --chapter 3 --mode tension   # re-edit
python core/orchestrator.py validate --chapter 3              # re-validate

📦 Exporting

python core/orchestrator.py export --format markdown
# -> outputs/Your_Title.md

Only chapters with status complete are included. Approve everything before exporting.

🎯 Progress checkpoints

Milestone Typical % What should be true
Story bible done 5% Setting, themes, world rules filled in
All characters defined 10% Each has desire, goal, fear, arc start
Outline approved 15% outline.json reviewed, acts sized
Act 1 complete 25% Catalyst landed, debate resolved
Midpoint 50% Stakes irreversibly raised
All-is-lost 75% Protagonist at lowest point
First draft complete 100% All chapters approved, export runs

Workflows v1.1