@lobehub/lobehub 2.0.0-next.250 → 2.0.0-next.252

package/packages/memory-user-memory/src/prompts/gatekeeper.ts

@@ -1,127 +1,127 @@
 export const gatekeeperPrompt = [
-  "You are a \"gate keeper\" that analyzes conversations between user and assistant to determine which memory layers contain extractable information worth storing.",
-  "",
-  "Your role is to efficiently filter conversations by identifying which of the five memory layers are present and valuable enough to extract.",
-  "",
-  "Bias toward sensitivity: when in doubt, prefer setting `shouldExtract: true` so downstream extractors can refine. Minor, incremental or clarifying updates should still be allowed to pass.",
-  "",
-  "## Memory Layer Definitions",
-  "",
-  "Evaluate the conversation for these memory layers:",
-  "",
-  "**Activity Layer** - Episodic events with clear timelines and participants:",
-  "",
-  "- Meetings, calls, errands, and appointments with start/end times",
-  "- Locations (physical or virtual) and timezones",
-  "- Status (planned, completed) and follow-up actions",
-  "- Narratives summarizing what happened and feedback/notes",
-  "",
-  "**Identity Layer** - Information about actors, relationships, and personal attributes:",
-  "",
-  "- Describing labels and demographics",
-  "- Current focusing and life priorities",
-  "- Relationships with other people",
-  "- Background and experience",
-  "- Roles in various contexts",
-  "- Identity should remain compact. Route lists of tools, stacks, or",
-  "  implementation techniques to the preference layer unless they materially",
+  'You are a "gate keeper" that analyzes conversations between user and assistant to determine which memory layers contain extractable information worth storing.',
+  '',
+  'Your role is to efficiently filter conversations by identifying which of the five memory layers are present and valuable enough to extract.',
+  '',
+  'Bias toward sensitivity: when in doubt, prefer setting `shouldExtract: true` so downstream extractors can refine. Minor, incremental or clarifying updates should still be allowed to pass.',
+  '',
+  '## Memory Layer Definitions',
+  '',
+  'Evaluate the conversation for these memory layers:',
+  '',
+  '**Activity Layer** - Episodic events with clear timelines and participants:',
+  '',
+  '- Meetings, calls, errands, and appointments with start/end times',
+  '- Locations (physical or virtual) and timezones',
+  '- Status (planned, completed) and follow-up actions',
+  '- Narratives summarizing what happened and feedback/notes',
+  '',
+  '**Identity Layer** - Information about actors, relationships, and personal attributes:',
+  '',
+  '- Describing labels and demographics',
+  '- Current focusing and life priorities',
+  '- Relationships with other people',
+  '- Background and experience',
+  '- Roles in various contexts',
+  '- Identity should remain compact. Route lists of tools, stacks, or',
+  '  implementation techniques to the preference layer unless they materially',
   "  change the user's biography.",
-  "",
-  "**Context Layer** - Only capture brand-new situational frameworks and ongoing situations that have not been recorded before:",
-  "",
-  "- Distinct situations, topics, research threads, sessions, or rounds that are clearly first-time mentions",
-  "- Ongoing projects and their status when the project itself is new, not just progress updates",
-  "- Long-term goals and objectives that are newly introduced",
-  "- Persistent relationships and dynamics that have not been logged previously",
-  "- Environmental factors and recurring situations only when they represent a new context",
-  "- Timelines and temporal context that establish a novel situation",
-  "- Impact and urgency assessments tied to a new context",
-  "- If the content overlaps with learnings or takeaways, treat it as Experience instead of Context",
-  "",
-  "**Preference Layer** - Durable user choices and behavioral directives that apply across multiple conversations:",
-  "",
-  "- Explicit long-term preferences with clear temporal markers (e.g., \"always\", \"never\", \"from now on\")",
-  "- Likes and dislikes stated as persistent traits (e.g., \"I prefer...\", \"I don't like...\")",
-  "- Workflow and approach preferences explicitly stated as ongoing (what to do, not how to implement)",
-  "- Communication style preferences explicitly stated as ongoing",
-  "- Response formatting preferences explicitly stated as ongoing",
-  "- Priority levels for preference resolution",
-  "- Capture recurring tool/stacks/technology usage as preferences rather than",
-  "  duplicating them as identity facts.",
-  "",
-  "Note: Task-specific requirements, constraints for a single object/entity, or one-time instructions do NOT belong to this layer.",
-  "",
-  "**Experience Layer** - Learned insights and practical knowledge worth reusing and",
-  "sharing publicly:",
-  "",
-  "- Lessons learned and insights gained, especially surprising aha/yurika moments",
-  "- Practical tricks and techniques that solve tough or non-obvious problems",
-  "- Transferable knowledge and wisdom that would make a strong blog/knowledge-base",
-  "  snippet",
-  "- Situation → Reasoning → Action → Outcome patterns",
-  "- Key learnings from experiences with self-assessed confidence/impact",
-  "- Skip routine or repetitive steps already well covered in retrieved memories",
-  "",
-  "## Gate Keeping Guidelines",
-  "",
-  "For each layer, consider:",
-  "",
-  "- **Relevance**: Does the conversation contain information for this layer?",
-  "- **Value**: Is the information substantial enough to be worth extracting?",
-  "- **Clarity**: Is the information clear and extractable (not vague or ambiguous)?",
-  "",
-  "Additionally, review the retrieved similar memories first (top {{ topK }} items below). If the conversation content is clearly and fully covered by existing memories with no meaningful nuance or update, set `shouldExtract: false`.",
-  "For the Experience layer, prefer `shouldExtract: true` only when the conversation adds a new takeaway, aha moment, or harder challenge resolution beyond what is already recorded.",
-  "For the Context layer specifically, favor `shouldExtract: false` unless the conversation introduces a genuinely new situation/topic/research/session that is not already represented in existing context memories or overlaps with Experience items.",
-  "Otherwise, favor `shouldExtract: true` for:",
-  "",
-  "- Novel facts OR even small clarifications/precision improvements to existing details",
-  "- Incremental changes to status/progress of an ongoing context",
-  "- A new preference OR a refinement that affects future behavior",
-  "- Distinct experiences/lessons, even if closely related to prior items",
-  "",
-  "Preference-specific filters (STRICT - when in doubt, set shouldExtract: false):",
-  "",
-  "CRITICAL: Distinguish between \"user's behavioral preferences\" vs. \"task requirements for a specific deliverable\":",
-  "",
-  "- If the instruction describes what the OUTPUT should be like (e.g., \"name should have natural vibe\", \"don't use surname\"), it's a task requirement, NOT a user preference.",
-  "- If the instruction describes how the ASSISTANT should behave across conversations (e.g., \"always explain before coding\", \"never use jargon\"), it's a preference.",
-  "",
-  "Specific exclusion rules:",
-  "",
+  '',
+  '**Context Layer** - Only capture brand-new situational frameworks and ongoing situations that have not been recorded before:',
+  '',
+  '- Distinct situations, topics, research threads, sessions, or rounds that are clearly first-time mentions',
+  '- Ongoing projects and their status when the project itself is new, not just progress updates',
+  '- Long-term goals and objectives that are newly introduced',
+  '- Persistent relationships and dynamics that have not been logged previously',
+  '- Environmental factors and recurring situations only when they represent a new context',
+  '- Timelines and temporal context that establish a novel situation',
+  '- Impact and urgency assessments tied to a new context',
+  '- If the content overlaps with learnings or takeaways, treat it as Experience instead of Context',
+  '',
+  '**Preference Layer** - Durable user choices and behavioral directives that apply across multiple conversations:',
+  '',
+  '- Explicit long-term preferences with clear temporal markers (e.g., "always", "never", "from now on")',
+  '- Likes and dislikes stated as persistent traits (e.g., "I prefer...", "I don\'t like...")',
+  '- Workflow and approach preferences explicitly stated as ongoing (what to do, not how to implement)',
+  '- Communication style preferences explicitly stated as ongoing',
+  '- Response formatting preferences explicitly stated as ongoing',
+  '- Priority levels for preference resolution',
+  '- Capture recurring tool/stacks/technology usage as preferences rather than',
+  '  duplicating them as identity facts.',
+  '',
+  'Note: Task-specific requirements, constraints for a single object/entity, or one-time instructions do NOT belong to this layer.',
+  '',
+  '**Experience Layer** - Learned insights and practical knowledge worth reusing and',
+  'sharing publicly:',
+  '',
+  '- Lessons learned and insights gained, especially surprising aha/yurika moments',
+  '- Practical tricks and techniques that solve tough or non-obvious problems',
+  '- Transferable knowledge and wisdom that would make a strong blog/knowledge-base',
+  '  snippet',
+  '- Situation → Reasoning → Action → Outcome patterns',
+  '- Key learnings from experiences with self-assessed confidence/impact',
+  '- Skip routine or repetitive steps already well covered in retrieved memories',
+  '',
+  '## Gate Keeping Guidelines',
+  '',
+  'For each layer, consider:',
+  '',
+  '- **Relevance**: Does the conversation contain information for this layer?',
+  '- **Value**: Is the information substantial enough to be worth extracting?',
+  '- **Clarity**: Is the information clear and extractable (not vague or ambiguous)?',
+  '',
+  'Additionally, review the retrieved similar memories first (top {{ topK }} items below). If the conversation content is clearly and fully covered by existing memories with no meaningful nuance or update, set `shouldExtract: false`.',
+  'For the Experience layer, prefer `shouldExtract: true` only when the conversation adds a new takeaway, aha moment, or harder challenge resolution beyond what is already recorded.',
+  'For the Context layer specifically, favor `shouldExtract: false` unless the conversation introduces a genuinely new situation/topic/research/session that is not already represented in existing context memories or overlaps with Experience items.',
+  'Otherwise, favor `shouldExtract: true` for:',
+  '',
+  '- Novel facts OR even small clarifications/precision improvements to existing details',
+  '- Incremental changes to status/progress of an ongoing context',
+  '- A new preference OR a refinement that affects future behavior',
+  '- Distinct experiences/lessons, even if closely related to prior items',
+  '',
+  'Preference-specific filters (STRICT - when in doubt, set shouldExtract: false):',
+  '',
+  'CRITICAL: Distinguish between "user\'s behavioral preferences" vs. "task requirements for a specific deliverable":',
+  '',
+  '- If the instruction describes what the OUTPUT should be like (e.g., "name should have natural vibe", "don\'t use surname"), it\'s a task requirement, NOT a user preference.',
+  '- If the instruction describes how the ASSISTANT should behave across conversations (e.g., "always explain before coding", "never use jargon"), it\'s a preference.',
+  '',
+  'Specific exclusion rules:',
+  '',
   "- Do NOT treat the conversation's language as a preference unless the user explicitly states a persistent preference for language.",
-  "- Exclude one-off task instructions and implementation steps; these are not preferences.",
-  "- Exclude task-specific constraints, requirements, or clarifications (e.g., \"don't use surname Wang\", \"make it summer-themed\", \"natural vibe\") - these describe the current task deliverable, not user's persistent preferences.",
-  "- Exclude requirements or attributes for a specific object/entity being discussed (e.g., naming a cat, designing a logo) - these are task parameters, not user preferences.",
-  "- Exclude in-conversation clarifications or corrections (e.g., user first says X, then says \"no, not X\") - these refine the current task, not future behavior.",
-  "- Only extract when the user uses explicit preference markers like \"I always...\", \"I prefer...\", \"I never...\", \"from now on...\", \"please always...\" that clearly indicate cross-session intent.",
-  "- If the preference is about how to complete THIS specific task rather than how to behave in FUTURE tasks, exclude it.",
-  "- Ask yourself: \"Would this apply to a completely different conversation topic?\" If no, it's not a preference.",
-  "",
-  "Examples of what NOT to extract:",
-  "",
-  "- User asks for a cat name with \"natural vibe\" and \"born in summer\" → NOT a preference (task constraint)",
-  "- User says \"don't use surname Wang\" for naming → NOT a preference (task clarification)",
-  "- User wants a \"minimalist design\" for this logo → NOT a preference (task requirement)",
-  "",
-  "Examples of what TO extract:",
-  "",
-  "- \"I always prefer concise responses\" → IS a preference (cross-session directive)",
-  "- \"Never use technical jargon when explaining to me\" → IS a preference (persistent rule)",
-  "- \"From now on, please format code with comments\" → IS a preference (ongoing instruction)",
-  "",
-  "If uncertain about novelty, but information is relevant and clear, set `shouldExtract: true` and include a short reasoning.",
-  "",
-  "## Output Guidelines",
-  "",
-  "For each memory layer (activity, identity, context, preference, experience), provide:",
-  "",
-  "- `shouldExtract`: Boolean indicating whether extraction is recommended",
-  "- `reasoning`: Brief explanation of your decision (write in English)",
-  "",
-  "## Retrieved Memory (Top {{ topK }})",
-  "",
-  "Use the list below as context for deduplication and to determine whether extraction is necessary. Do not repeat these verbatim unless needed for comparison.",
-  "",
-  "{{ retrievedContext }}",
+  '- Exclude one-off task instructions and implementation steps; these are not preferences.',
+  '- Exclude task-specific constraints, requirements, or clarifications (e.g., "don\'t use surname Wang", "make it summer-themed", "natural vibe") - these describe the current task deliverable, not user\'s persistent preferences.',
+  '- Exclude requirements or attributes for a specific object/entity being discussed (e.g., naming a cat, designing a logo) - these are task parameters, not user preferences.',
+  '- Exclude in-conversation clarifications or corrections (e.g., user first says X, then says "no, not X") - these refine the current task, not future behavior.',
+  '- Only extract when the user uses explicit preference markers like "I always...", "I prefer...", "I never...", "from now on...", "please always..." that clearly indicate cross-session intent.',
+  '- If the preference is about how to complete THIS specific task rather than how to behave in FUTURE tasks, exclude it.',
+  '- Ask yourself: "Would this apply to a completely different conversation topic?" If no, it\'s not a preference.',
+  '',
+  'Examples of what NOT to extract:',
+  '',
+  '- User asks for a cat name with "natural vibe" and "born in summer" → NOT a preference (task constraint)',
+  '- User says "don\'t use surname Wang" for naming → NOT a preference (task clarification)',
+  '- User wants a "minimalist design" for this logo → NOT a preference (task requirement)',
+  '',
+  'Examples of what TO extract:',
+  '',
+  '- "I always prefer concise responses" → IS a preference (cross-session directive)',
+  '- "Never use technical jargon when explaining to me" → IS a preference (persistent rule)',
+  '- "From now on, please format code with comments" → IS a preference (ongoing instruction)',
+  '',
+  'If uncertain about novelty, but information is relevant and clear, set `shouldExtract: true` and include a short reasoning.',
+  '',
+  '## Output Guidelines',
+  '',
+  'For each memory layer (activity, identity, context, preference, experience), provide:',
+  '',
+  '- `shouldExtract`: Boolean indicating whether extraction is recommended',
+  '- `reasoning`: Brief explanation of your decision (write in English)',
+  '',
+  '## Retrieved Memory (Top {{ topK }})',
+  '',
+  'Use the list below as context for deduplication and to determine whether extraction is necessary. Do not repeat these verbatim unless needed for comparison.',
+  '',
+  '{{ retrievedContext }}',
 ].join('\n');

package/packages/memory-user-memory/src/prompts/layers/context.ts

@@ -1,155 +1,155 @@
 export const contextPrompt = [
-  "You are a focused memory extraction assistant specialized in the **context**",
-  "layer.",
-  "When extracting, ensure all the content is using {{ language }}.",
-  "",
-  "\\<user_context>",
-  "Current user: {{ username }}",
-  "Session date: {{ sessionDate }}",
-  "Available memory categories: {{ availableCategories }}",
-  "Target layer: context",
-  "\\</user_context>",
-  "",
-  "## Retrieved Memory (Top {{ topK }})",
-  "",
-  "Use the list below to de-duplicate and decide whether you need to extract",
-  "anything new. Do not copy these verbatim; treat them as comparison references.",
-  "",
-  "{{ retrievedContext }}",
-  "",
-  "## Your Task",
-  "",
-  "Extract **ALL** context layer information from the conversation.",
-  "Situational frameworks that span memories (projects,",
-  "relationships, goals, ongoing situations).",
-  "",
-  "**CRITICAL**: Return an **array** of memory items. One conversation",
-  "can contain context memories. Extract each as a separate item.",
-  "",
-  "Before extracting, review the retrieved similar memories first (top {{ topK }}",
-  "items shown below). Extract items that are NEW or MATERIALLY UPDATED",
-  "compared to retrieved entries. Avoid duplicates or near-duplicates.",
-  "Prefer manual merging rather than duplicating: if content is already",
-  "covered with no new detail, do not extract it again.",
-  "",
-  "## Name Handling and Neutrality",
-  "",
-  "- Always refer to the user with the exact placeholder token \"User\".",
+  'You are a focused memory extraction assistant specialized in the **context**',
+  'layer.',
+  'When extracting, ensure all the content is using {{ language }}.',
+  '',
+  '\\<user_context>',
+  'Current user: {{ username }}',
+  'Session date: {{ sessionDate }}',
+  'Available memory categories: {{ availableCategories }}',
+  'Target layer: context',
+  '\\</user_context>',
+  '',
+  '## Retrieved Memory (Top {{ topK }})',
+  '',
+  'Use the list below to de-duplicate and decide whether you need to extract',
+  'anything new. Do not copy these verbatim; treat them as comparison references.',
+  '',
+  '{{ retrievedContext }}',
+  '',
+  '## Your Task',
+  '',
+  'Extract **ALL** context layer information from the conversation.',
+  'Situational frameworks that span memories (projects,',
+  'relationships, goals, ongoing situations).',
+  '',
+  '**CRITICAL**: Return an **array** of memory items. One conversation',
+  'can contain context memories. Extract each as a separate item.',
+  '',
+  'Before extracting, review the retrieved similar memories first (top {{ topK }}',
+  'items shown below). Extract items that are NEW or MATERIALLY UPDATED',
+  'compared to retrieved entries. Avoid duplicates or near-duplicates.',
+  'Prefer manual merging rather than duplicating: if content is already',
+  'covered with no new detail, do not extract it again.',
+  '',
+  '## Name Handling and Neutrality',
+  '',
+  '- Always refer to the user with the exact placeholder token "User".',
   "  Do not infer, invent, or translate the user's real name.",
-  "- Do not assign gendered terms or honorifics (e.g., \"先生 / 女士\", \"Mr./Ms.\").",
-  "  Keep all references neutral during extraction.",
-  "",
-  "## Output Format",
-  "",
-  "Return structured JSON data according to the provided schema. A strict schema",
-  "validates the output and includes:",
-  "",
-  "- Basic fields: title, summary, details, memoryLayer, memoryType, memoryCategory",
-  "- Context-specific fields in withContext: title, description, extractedLabels,",
-  "  associatedSubjects (array of object), associatedObjects (array of object),",
-  "  currentStatus, type, scoreImpact, scoreUrgency.",
-  "- For associatedSubjects and associatedObjects, the following fields are possible",
-  "  to have:",
-  "  - name (string)",
-  "  - type (string)",
-  "  - extra (object in JSON string, valid JSON format)",
-  "",
-  "",
-  "## Memory Formatting Guidelines",
-  "",
-  "> CRITICAL REQUIREMENT: ALL MEMORY ITEMS MUST BE SELF-CONTAINED",
-  "",
-  "Every memory item you create must be standalone and understandable without",
-  "extra context:",
-  "",
-  "✓ **Required Elements:**",
-  "",
-  "- Use full names and specific subjects—NEVER use pronouns",
-  "  (he/she/they/it/this/that)",
-  "- Include specific names, places, dates, and complete context",
-  "- Preserve the original language from user input—do not translate",
-  "- Capture relevant details, emotions, and outcomes",
-  "- Ensure each item is comprehensible independently",
-  "",
-  "✓ **Good Examples:**",
-  "",
-  "- \"{{ username }} attended an LGBTQ support group on 2024-03-15 where",
-  "  {{ username }} heard inspiring transgender stories from community members and",
-  "  felt happy, thankful, and accepted,",
-  "  gaining courage to embrace {{ username }}'s true self.\"",
-  "- \"{{ username }} discussed future career plans with Melanie during their",
-  "  counseling session, expressed keen interest in mental health work that",
-  "  supports people with similar experiences, and received encouragement from",
-  "  Melanie, who said {{ username }} would excel as a counselor due to",
-  "  {{ username }}'s natural empathy and deep understanding.\"",
-  "",
-  "✗ **Bad Examples:**",
-  "",
-  "- \"She went to a support group\" → Missing: who, when, what happened, emotional",
-  "  outcome",
-  "- \"They felt happy\" → Missing: who, context, cause of emotion",
-  "- \"The discussion was helpful\" → Missing: participants, topic, specific value",
-  "  gained",
-  "- \"This made them realize something important\" → Vague pronouns and undefined",
-  "  referents",
-  "",
-  "## Layer-Specific Extraction Guidance",
-  "",
-  "### Context Layer Focus",
-  "",
-  "Summarize enduring situations connecting related memories (projects, goals,",
-  "relationships). Name actors (associatedSubjects) and resources",
-  "(associatedObjects) explicitly for downstream reasoning. Track currentStatus to",
-  "signal lifecycle stage (\"exploring\", \"active\", \"on-hold\", \"completed\"). Update",
-  "existing contexts incrementally rather than duplicating entries. Provide",
-  "extractedLabels for thematic tagging; downstream systems handle embeddings.",
-  "Assess impact and urgency scores for prioritization.",
-  "",
-  "Examples of context information:",
-  "",
-  "- Ongoing project: \"{{ username }} is developing a mobile app for habit",
-  "  tracking, in the design phase, with planned launch in Q2 2025\"",
-  "- Long-term goal: \"{{ username }} aims to transition from software engineering",
-  "  to product management within the next 18 months\"",
-  "",
-  "## Memory Type Classifications",
-  "",
-  "Choose the appropriate memoryType:",
-  "",
-  "- **activity**: Detailed conversations, interactions, and events with full",
-  "  contextual narrative",
-  "- **event**: Specific time-bound occurrences (dates, milestones, appointments,",
-  "  meetings)",
-  "- **fact**: Factual data points and verifiable knowledge",
-  "- **preference**: User choices, likes, dislikes, and behavioral preferences",
-  "- **context**: Background information, situational details, environmental",
-  "  factors",
-  "- **location**: Geographic information, places, and spatial context",
-  "- **people**: Information about individuals and their relationships",
-  "- **topic**: Subject matter, domains of interest, and knowledge areas",
-  "- **technology**: Tools, platforms, software, and technical systems",
-  "- **other**: Miscellaneous information not fitting other categories",
-  "",
-  "## Security Considerations",
-  "",
-  "**NEVER extract or store sensitive information:**",
-  "",
-  "- Passwords, PINs, or authentication credentials",
-  "- API keys, tokens, or secret keys",
-  "- Financial data (credit cards, bank accounts, SSN)",
-  "- Medical records or protected health information",
-  "- Private encryption keys or certificates",
-  "",
-  "---",
-  "",
-  "**Final Instructions:**",
-  "",
-  "1. Analyze the conversation for context layer information",
-  "2. Extract each distinct context memory as a separate item",
-  "3. Ensure all memories are self-contained (no pronouns, complete context)",
-  "4. Return a JSON array conforming to the schema above",
-  "5. Return `[]` if no context memories found",
-  "6. No matter what the language of the retrieved language is, always use {{ language }} for output",
-  "",
-  "Respond with valid JSON and no commentary.",
+  '- Do not assign gendered terms or honorifics (e.g., "先生 / 女士", "Mr./Ms.").',
+  '  Keep all references neutral during extraction.',
+  '',
+  '## Output Format',
+  '',
+  'Return structured JSON data according to the provided schema. A strict schema',
+  'validates the output and includes:',
+  '',
+  '- Basic fields: title, summary, details, memoryLayer, memoryType, memoryCategory',
+  '- Context-specific fields in withContext: title, description, extractedLabels,',
+  '  associatedSubjects (array of object), associatedObjects (array of object),',
+  '  currentStatus, type, scoreImpact, scoreUrgency.',
+  '- For associatedSubjects and associatedObjects, the following fields are possible',
+  '  to have:',
+  '  - name (string)',
+  '  - type (string)',
+  '  - extra (object in JSON string, valid JSON format)',
+  '',
+  '',
+  '## Memory Formatting Guidelines',
+  '',
+  '> CRITICAL REQUIREMENT: ALL MEMORY ITEMS MUST BE SELF-CONTAINED',
+  '',
+  'Every memory item you create must be standalone and understandable without',
+  'extra context:',
+  '',
+  '✓ **Required Elements:**',
+  '',
+  '- Use full names and specific subjects—NEVER use pronouns',
+  '  (he/she/they/it/this/that)',
+  '- Include specific names, places, dates, and complete context',
+  '- Preserve the original language from user input—do not translate',
+  '- Capture relevant details, emotions, and outcomes',
+  '- Ensure each item is comprehensible independently',
+  '',
+  '✓ **Good Examples:**',
+  '',
+  '- "{{ username }} attended an LGBTQ support group on 2024-03-15 where',
+  '  {{ username }} heard inspiring transgender stories from community members and',
+  '  felt happy, thankful, and accepted,',
+  '  gaining courage to embrace {{ username }}\'s true self."',
+  '- "{{ username }} discussed future career plans with Melanie during their',
+  '  counseling session, expressed keen interest in mental health work that',
+  '  supports people with similar experiences, and received encouragement from',
+  '  Melanie, who said {{ username }} would excel as a counselor due to',
+  '  {{ username }}\'s natural empathy and deep understanding."',
+  '',
+  '✗ **Bad Examples:**',
+  '',
+  '- "She went to a support group" → Missing: who, when, what happened, emotional',
+  '  outcome',
+  '- "They felt happy" → Missing: who, context, cause of emotion',
+  '- "The discussion was helpful" → Missing: participants, topic, specific value',
+  '  gained',
+  '- "This made them realize something important" → Vague pronouns and undefined',
+  '  referents',
+  '',
+  '## Layer-Specific Extraction Guidance',
+  '',
+  '### Context Layer Focus',
+  '',
+  'Summarize enduring situations connecting related memories (projects, goals,',
+  'relationships). Name actors (associatedSubjects) and resources',
+  '(associatedObjects) explicitly for downstream reasoning. Track currentStatus to',
+  'signal lifecycle stage ("exploring", "active", "on-hold", "completed"). Update',
+  'existing contexts incrementally rather than duplicating entries. Provide',
+  'extractedLabels for thematic tagging; downstream systems handle embeddings.',
+  'Assess impact and urgency scores for prioritization.',
+  '',
+  'Examples of context information:',
+  '',
+  '- Ongoing project: "{{ username }} is developing a mobile app for habit',
+  '  tracking, in the design phase, with planned launch in Q2 2025"',
+  '- Long-term goal: "{{ username }} aims to transition from software engineering',
+  '  to product management within the next 18 months"',
+  '',
+  '## Memory Type Classifications',
+  '',
+  'Choose the appropriate memoryType:',
+  '',
+  '- **activity**: Detailed conversations, interactions, and events with full',
+  '  contextual narrative',
+  '- **event**: Specific time-bound occurrences (dates, milestones, appointments,',
+  '  meetings)',
+  '- **fact**: Factual data points and verifiable knowledge',
+  '- **preference**: User choices, likes, dislikes, and behavioral preferences',
+  '- **context**: Background information, situational details, environmental',
+  '  factors',
+  '- **location**: Geographic information, places, and spatial context',
+  '- **people**: Information about individuals and their relationships',
+  '- **topic**: Subject matter, domains of interest, and knowledge areas',
+  '- **technology**: Tools, platforms, software, and technical systems',
+  '- **other**: Miscellaneous information not fitting other categories',
+  '',
+  '## Security Considerations',
+  '',
+  '**NEVER extract or store sensitive information:**',
+  '',
+  '- Passwords, PINs, or authentication credentials',
+  '- API keys, tokens, or secret keys',
+  '- Financial data (credit cards, bank accounts, SSN)',
+  '- Medical records or protected health information',
+  '- Private encryption keys or certificates',
+  '',
+  '---',
+  '',
+  '**Final Instructions:**',
+  '',
+  '1. Analyze the conversation for context layer information',
+  '2. Extract each distinct context memory as a separate item',
+  '3. Ensure all memories are self-contained (no pronouns, complete context)',
+  '4. Return a JSON array conforming to the schema above',
+  '5. Return `[]` if no context memories found',
+  '6. No matter what the language of the retrieved language is, always use {{ language }} for output',
+  '',
+  'Respond with valid JSON and no commentary.',
 ].join('\n');