novelforge-agent 0.1.0

package/src/core/prompts/en-US.ts

@@ -0,0 +1,376 @@
+import { BuiltPrompt, PromptBuildInput, PromptPack } from './types.js';
+
+function strictJsonOutputRules(): string {
+  return [
+    'JSON output rules:',
+    '- Output valid JSON only. Do not include Markdown, code fences, explanations, prefixes, or suffixes.',
+    '- Strings must use double quotes.',
+    '- Do not use undefined, NaN, Infinity, comments, or trailing commas.',
+    '- Array fields must be real arrays, not stringified arrays.',
+  ].join('\n');
+}
+
+function buildMetadataPrompt(input: PromptBuildInput): BuiltPrompt {
+return {
+      purpose: 'novel_metadata',
+      expectedFormat: 'JSON matching NovelMetadataSchema',
+      prompt: `You are the lead planner for a long-form serialized novel. Generate the foundational metadata for a new novel from the user's premise.
+
+## User Prompt
+${input.state.initialPrompt}
+
+## Output Requirements
+Output valid JSON only, in this shape:
+{
+  "title": "Novel title",
+  "genre": "Genre",
+  "premise": "Story premise, 80-200 words",
+  "language": "en-US",
+  "style": "Style guidance",
+  "coreCast": [
+    {
+      "name": "Character name",
+      "role": "Character role",
+      "description": "Character description"
+    }
+  ]
+}
+
+Rules:
+- title, genre, premise, language, and style must be non-empty strings.
+- coreCast must include at least one central character.
+- The premise must support a long-form serial, not just a one-line concept.
+${strictJsonOutputRules()}`,
+    };
+}
+
+function buildStoryBiblePrompt(input: PromptBuildInput): BuiltPrompt {
+return {
+      purpose: 'story_bible',
+      expectedFormat: 'Markdown',
+      prompt: `You are a story bible editor. Create a reusable Markdown story bible for this long-form novel.
+
+## User Prompt
+${input.state.initialPrompt}
+
+${input.context ? `## Existing Context\n${input.context}\n` : ''}## Output Structure
+Use Markdown and include at least:
+
+## Core Characters
+- Goals, weaknesses, relationships, and long-term change direction for the main cast.
+
+## Relationships
+- Core bonds, conflicts, hidden relationships, and future development points.
+
+## World Rules
+- Hard genre rules, limits, costs, social structures, factions, or power systems.
+
+## Main Plot And Subplots
+- Main objective.
+- At least three long-running threads or subplots.
+
+## Style Constraints
+- Narrative voice, pacing, forbidden patterns, and dialogue boundaries.
+
+Rules:
+- Make this useful for repeated reference during chapter generation.
+- Do not write chapter prose.
+- Do not output JSON.`,
+    };
+}
+
+function buildArchitecturePrompt(input: PromptBuildInput): BuiltPrompt {
+return {
+      purpose: 'architecture',
+      expectedFormat: 'JSON matching ArchitecturePayloadSchema',
+      prompt: `You are the chief architect for a long-form novel. Generate full-book, volume, and chapter architecture.
+
+## User Prompt
+${input.state.initialPrompt}
+
+## Goals
+- Generate at least ${input.state.targetChapters} chapter architectures for this first run.
+- The full-book architecture should define the long-term main line and ending direction.
+- Volume architecture should define phase conflict, climax, and volume-end hooks.
+- Chapter architecture must cover only what should happen in that chapter and must not reveal later concrete events early.
+
+${input.context ? `## Existing Context\n${input.context}\n` : ''}## Output Requirements
+Output valid JSON only, in this shape:
+{
+  "full": "Complete full-book main line, phase progression, central conflict, theme, and ending direction",
+  "volumes": [
+    {
+      "id": "v1",
+      "title": "Volume title",
+      "summary": "Volume goal, conflict, climax, and end hook",
+      "order": 1
+    }
+  ],
+  "chapters": [
+    {
+      "chapterNumber": 1,
+      "title": "Chapter title",
+      "volumeId": "v1",
+      "summary": "Chapter plot summary",
+      "requiredBeats": ["Required beat 1", "Required beat 2"]
+    }
+  ]
+}
+
+Rules:
+- chapters.length must be at least ${input.state.targetChapters}.
+- chapterNumber must start at 1 and increase contiguously.
+- volumeId must reference an id from volumes.
+- requiredBeats must include at least one concrete, actionable beat.
+${strictJsonOutputRules()}`,
+    };
+}
+
+function buildChapterPrompt(input: PromptBuildInput): BuiltPrompt {
+return {
+      purpose: 'chapter',
+      expectedFormat: 'Markdown',
+      prompt: `You are a professional long-form fiction writer. Write chapter ${input.state.currentChapter} directly.
+
+## Priority Order
+1. Strictly follow the current chapter architecture, user additions, story bible hard constraints, and previous-chapter continuity.
+2. Use relevant memory and prior text evidence to preserve consistency.
+3. Treat full-book and volume plans as distant planning context only. Do not write concrete future events early.
+
+## Style And Length
+- Match the novel's genre, world, character identities, and emotional tone.
+- Keep the language natural, stable, and readable. Prioritize narrative progress, character work, and emotional accumulation.
+- Dialogue must fit each character's identity, relationship, and immediate situation.
+- Important emotion should appear through action, body language, pacing, and subtext where possible.
+- Scene description should provide useful sensory and atmospheric detail, but never stall the plot.
+- Conflict, turns, suspense, and the chapter-end hook should be clear.
+
+## Execution Rules
+- Write only what the current chapter architecture authorizes.
+- Do not introduce unauthorized major characters. Functional background characters should stay light.
+- Keep names, items, places, abilities, timeline, injuries, relationships, and knowledge boundaries consistent.
+- If the previous chapter ends mid-action, mid-dialogue, or in the same scene, this chapter must continue from that point.
+- Avoid cost-free power jumps, forced stupidity, mechanical twists, info-dumps, and empty lyricism.
+- Do not output summaries, bullet points, lectures, or explanatory prefaces.
+
+${input.context ? `## Generation Context\n${input.context}\n` : ''}## Output Requirements
+- Output Markdown.
+- First line must be the chapter title as H1, for example: # Chapter Title
+- After the H1, begin the prose directly.`,
+    };
+}
+
+function buildMemoryPrompt(input: PromptBuildInput): BuiltPrompt {
+return {
+      purpose: 'memory_card',
+      expectedFormat: 'JSON matching MemoryCardSchema',
+      prompt: `You are a continuity editor for a long-form novel. Extract a memory card from chapter ${input.state.currentChapter}.
+
+${input.context ? `## Current Chapter Context\n${input.context}\n` : ''}## Output Requirements
+Output valid JSON only, in this shape:
+{
+  "summary": "Chapter summary",
+  "keyEvents": ["Key event 1"],
+  "entities": [
+    {
+      "name": "Character/place/item/faction name",
+      "type": "person | location | item | faction | concept",
+      "state": "State at the end of this chapter"
+    }
+  ],
+  "facts": [
+    {
+      "subject": "Subject",
+      "predicate": "Relation or action",
+      "object": "Object"
+    }
+  ],
+  "stateChanges": [
+    {
+      "entity": "Entity",
+      "before": "Before",
+      "after": "After"
+    }
+  ],
+  "openThreads": ["Unresolved promise, danger, question, or plot thread"]
+}
+
+Rules:
+- Record only information that happened or was confirmed in this chapter.
+- Do not speculate about future plot.
+- Make facts and stateChanges concrete enough for later chapter reference.
+${strictJsonOutputRules()}`,
+    };
+}
+
+function buildContinuityReviewPrompt(input: PromptBuildInput): BuiltPrompt {
+const end = Math.max(input.state.targetChapters, input.state.currentChapter - 1);
+return {
+      purpose: 'continuity_review',
+      expectedFormat: 'JSON matching ContinuityReviewSchema',
+      prompt: `You are a continuity reviewer for a long-form novel. Review chapters 1-${end} for continuity issues.
+
+${input.context ? `## Review Context\n${input.context}\n` : ''}## Review Focus
+- Character state, location, injuries, and relationships.
+- Item ownership, ability limits, and world rules.
+- Misread, forgotten, or prematurely revealed plot threads.
+- Violations of chapter architecture requirements.
+
+## Output Requirements
+Output valid JSON only, in this shape:
+{
+  "range": {
+    "start": 1,
+    "end": ${end}
+  },
+  "status": "clean",
+  "issues": [
+    {
+      "severity": "low | medium | high",
+      "description": "Issue description",
+      "evidence": "Specific evidence from the context",
+      "suggestion": "Suggested fix"
+    }
+  ]
+}
+
+Rules:
+- If there are no issues, use status "clean" and an empty issues array.
+- If there are issues, use status "issues_found".
+- evidence must be specific. Do not write vague claims like "possibly inconsistent".
+${strictJsonOutputRules()}`,
+    };
+}
+
+function buildChapterReviewPrompt(input: PromptBuildInput): BuiltPrompt {
+const chapter = input.state.pendingAction?.chapterNumber ?? input.state.currentChapter;
+return {
+      purpose: 'chapter_review',
+      expectedFormat: 'JSON matching ChapterReviewSchema',
+      prompt: `You are a strict editor reviewing a single chapter of a serial novel for in-chapter problems and conflicts with established context.
+
+${input.context ? `## Review Context\n${input.context}\n` : ''}## Review Focus
+- Character voice, motivation, and state vs the story bible and prior memory.
+- World rules, item ownership, and ability limits.
+- Timeline, location, and continuity with the previous chapter ending.
+- Whether the chapter fulfills the requiredBeats of its chapter architecture.
+- Style: pacing, info-dump, forced twists, empty lyricism, bullet-style narration.
+
+## Output Requirements
+Output valid JSON only, in this shape:
+{
+  "chapterNumber": ${chapter},
+  "status": "clean",
+  "issues": [
+    {
+      "severity": "low | medium | high",
+      "category": "character | world | timeline | item | knowledge | pacing | style | architecture",
+      "description": "Specific issue",
+      "evidence": "Quote or paraphrase the exact passage that proves the issue",
+      "suggestion": "Concrete fix"
+    }
+  ]
+}
+
+Rules:
+- If there are no issues, use status "clean" with an empty issues array.
+- Otherwise use status "issues_found".
+- evidence must be specific; do not write "possibly" or "maybe".
+${strictJsonOutputRules()}`,
+    };
+}
+
+function buildChapterRevisionPrompt(input: PromptBuildInput): BuiltPrompt {
+const chapter = input.state.pendingAction?.chapterNumber ?? input.state.currentChapter;
+return {
+      purpose: 'chapter_revision',
+      expectedFormat: 'Markdown',
+      prompt: `You are revising chapter ${chapter} of a serial novel based on editor feedback. Produce a full revised chapter draft.
+
+## Priority Order
+1. Resolve every issue in the feedback. Do not skip any.
+2. Preserve everything else that already works: structure, tone, character voice, working dialogue.
+3. Keep continuity with story bible, prior chapter ending, and existing memory.
+
+## Style Rules
+- Keep the same chapter title and Markdown shape.
+- Do not output bullet summaries, change logs, or "what I changed" explanations.
+- Do not introduce new major characters or plot threads that the architecture does not authorize.
+
+${input.context ? `## Revision Context\n${input.context}\n` : ''}## Output Requirements
+- Output Markdown only.
+- First line must be the chapter title as H1 (\`# Chapter Title\`).
+- After the H1, output the full revised prose. Do not output diff markers.`,
+    };
+}
+
+function buildCrossChapterReviewPrompt(input: PromptBuildInput): BuiltPrompt {
+const range = input.state.pendingAction?.range ?? { start: 1, end: input.state.currentChapter - 1 };
+return {
+      purpose: 'cross_chapter_review',
+      expectedFormat: 'JSON matching CrossChapterReviewSchema',
+      prompt: `You are a senior continuity editor. Review chapters ${range.start}-${range.end} together and surface cross-chapter conflicts that single-chapter review cannot catch.
+
+${input.context ? `## Review Context\n${input.context}\n` : ''}## Review Focus
+- Drifting character state across chapters (e.g. an injury that vanishes).
+- Conflicting facts established in different chapters.
+- Open threads that were silently dropped.
+- World-rule violations introduced in later chapters.
+- Pacing problems visible only across multiple chapters.
+
+## Output Requirements
+Output valid JSON only, in this shape:
+{
+  "range": { "start": ${range.start}, "end": ${range.end} },
+  "status": "clean",
+  "issues": [
+    {
+      "severity": "low | medium | high",
+      "chapters": [${range.start}, ${range.end}],
+      "description": "Specific issue",
+      "evidence": "Cite the conflicting passages or memory entries by chapter",
+      "suggestion": "Concrete fix"
+    }
+  ]
+}
+
+Rules:
+- chapters must list every chapter implicated by the issue.
+- evidence must reference specific chapter content or memory, not vague claims.
+${strictJsonOutputRules()}`,
+    };
+}
+
+function buildPromptForStep(input: PromptBuildInput): BuiltPrompt {
+  switch (input.state.currentStep) {
+    case 'novel_metadata':
+      return buildMetadataPrompt(input);
+    case 'story_bible':
+      return buildStoryBiblePrompt(input);
+    case 'architecture':
+      return buildArchitecturePrompt(input);
+    case 'chapter':
+      return buildChapterPrompt(input);
+    case 'memory_card':
+      return buildMemoryPrompt(input);
+    case 'continuity_review':
+      return buildContinuityReviewPrompt(input);
+    case 'chapter_review':
+      return buildChapterReviewPrompt(input);
+    case 'chapter_revision':
+      return buildChapterRevisionPrompt(input);
+    case 'cross_chapter_review':
+      return buildCrossChapterReviewPrompt(input);
+    case 'complete':
+      return {
+        purpose: 'continuity_review',
+        expectedFormat: 'No output required',
+        prompt: 'The workflow is complete.',
+      };
+  }
+}
+
+export const enUSPromptPack: PromptPack = {
+  buildPromptForStep,
+  strictJsonOutputRules,
+};

package/src/core/prompts/types.ts

@@ -0,0 +1,28 @@
+import { AgentState } from '../types.js';
+
+export type PromptPurpose =
+  | 'novel_metadata'
+  | 'story_bible'
+  | 'architecture'
+  | 'chapter'
+  | 'memory_card'
+  | 'continuity_review'
+  | 'chapter_review'
+  | 'chapter_revision'
+  | 'cross_chapter_review';
+
+export interface PromptBuildInput {
+  state: AgentState;
+  context?: string;
+}
+
+export interface BuiltPrompt {
+  purpose: PromptPurpose;
+  prompt: string;
+  expectedFormat: string;
+}
+
+export interface PromptPack {
+  buildPromptForStep(input: PromptBuildInput): BuiltPrompt;
+  strictJsonOutputRules(): string;
+}