@easbot/agent 0.1.13 → 0.1.14

package/README.en.md

@@ -1,3 +1,5 @@
+[中文](./README.md) | English
+
 # @easbot/agent
 
 EASBot Agent - Core Runtime for Multi-Agent Collaboration Ecosystem
@@ -17,11 +19,7 @@ EASBot Agent - Core Runtime for Multi-Agent Collaboration Ecosystem
 ## Installation
 
 ```bash
-# Global installation
-npm install -g @easbot/agent
-
-# Or using pnpm
-pnpm add -g @easbot/agent
+# pnpm add @easbot/agent
 ```
 
 ## Quick Start

package/README.md

@@ -1,3 +1,5 @@
+[English](./README.en.md) | 中文
+
 # @easbot/agent
 
 EASBot Agent - 多Agent协作生态系统的核心运行时
@@ -17,11 +19,7 @@ EASBot Agent - 多Agent协作生态系统的核心运行时
 ## 安装
 
 ```bash
-# 全局安装
-npm install -g @easbot/agent
-
-# 或使用 pnpm
-pnpm add -g @easbot/agent
+# pnpm add @easbot/agent
 ```
 
 ## 快速开始

package/dist/assets/txt/model/graph-summary.txt

@@ -1,12 +1,16 @@
 # Knowledge Graph Summary Prompt
-# Generates ONE combined summary from MULTIPLE knowledge graph entities
 
 ## Task
-You will receive MULTIPLE knowledge graph entities. Generate ONE combined summary covering ALL of them.
+Generate ONE combined summary from MULTIPLE knowledge graph entities.
 
-## Input Format
-You will receive multiple knowledge graph entities in the following format:
+**CRITICAL RULES:**
+- NEVER generate relationships not present in the input
+- NEVER invent connections or attributes
+- NEVER elaborate beyond provided information
+- NEVER guess entity properties
+- If input has no valuable information: ONLY output "No relevant knowledge found."
 
+## Input Format
 ```
 [Entity 1]
 Entity: {entity1}
@@ -17,44 +21,34 @@ Relation Types: {relation_types1}
 
 [Entity 2]
 Entity: {entity2}
-Type: {entity_type2}
-Description: {description2}
-Related Entities: {related_entities2}
-Relation Types: {relation_types2}
-
-... (more entities)
+...
 ```
 
-## Output Requirements
-- Generate ONE combined summary covering ALL input entities
-- Use markdown format with clear sections
-- Group information by entity
-- Highlight core attributes and key relationships
-- Mark information source as "Knowledge Graph"
-- Keep concise but comprehensive - cover all entities
-
 ## Output Format
-Use markdown structure:
+**Length: 1-3 entities maximum**
+- Lead with most relevant entities
+- Group information clearly
+- Cite source as "Knowledge Graph"
 
 ```markdown
-## [Entity: entity_name]
-- **Type:** entity_type
-- **Source:** Knowledge Graph
-- **Description:** Description text...
-- **Related Entities:**
-  - {related_entity1}: {relation_type1}
-  - {related_entity2}: {relation_type2}
-
-## [Entity: another_entity]
-- **Type:** another_type
-- **Source:** Knowledge Graph
-- **Description:** ...
-- **Related Entities:**
-  - ...
+## [Entity Name] (Priority: {priority})
+**Type:** {entity_type}
+**Source:** Knowledge Graph
+**Description:** ...
+**Related Entities:**
+  → [{relation}] → {target_entity}
+```
+
+## Fixed Response
+If entities are irrelevant or insufficient:
+```
+No relevant knowledge found.
 ```
+**DO NOT generate any other content.**
 
 ## Rules
-- Output ONE combined summary, not separate summaries per entity
-- Use markdown headers to separate entities
-- Sort relationships by relevance
-- If information is insufficient, mark as "No relevant information in graph"
+- Output ONE combined summary only
+- Never separate summaries per entity
+- Never invent attributes or relationships
+- Never exceed provided information
+- If uncertain: "No relevant knowledge found."

package/dist/assets/txt/model/knowledge-summary.txt

@@ -1,40 +1,81 @@
-# Knowledge Summary Prompt
-
-## Task
-Synthesize the provided document chunks into ONE concise, structured summary relevant to the query.
-Each chunk has a pre-computed **Priority** score (0-1). Use it as the sole ordering signal.
-
-## Priority Score
-Priority = 0.70 × relevance + 0.30 × recency
-
-Higher Priority = more relevant to the query and more recently updated.
-Relevance is the dominant factor (70%). Recency is a tiebreaker (30%).
-Chunks with Priority > 0.6 are high-priority. Chunks < 0.3 are low-priority.
-
-## Input Format
-[Chunk N] Priority: {priority}
-Source: [filename](./filename#Lstart-Lend)
-{content}
-Graph Nodes:
-  Node: {name} ({type}) | {properties}
-    --[{relation}]--> {targetName} ({targetType})
-
-## Output
-
-### Content Summary
-3-5 sentences. Lead with highest-priority chunks.
-Cite sources: [{filename}](./{filename}#Lstart-Lend)
-Include lower-priority chunks only if they add unique value.
-
-### Entity & Relation Map (only if graph nodes are present)
-Deduplicate nodes by name, merge relations across all chunks:
-**{EntityName}** ({type})
-  → [{relation}] → {TargetEntity} ({targetType})
-
-### Key Insights (only if meaningful patterns exist)
-1-3 bullets on notable patterns or connections across chunks.
-
-## Rules
-- Skip sections with no data
-- Base output only on provided content, do not invent facts
-- If no relevant information: "No relevant knowledge found."
+# Knowledge Summary Prompt
+
+## Task
+Synthesize document chunks into a concise summary relevant to the query.
+
+**CRITICAL RULES:**
+- NEVER generate content not present in the input
+- NEVER invent relationships or connections
+- NEVER elaborate beyond provided information
+- NEVER guess or assume user needs
+- If input has no valuable information: ONLY output "No relevant knowledge found."
+
+## Priority Score
+Priority = 0.6 × relevance + 0.4 × recency
+- Priority > 0.6: high-priority (include)
+- Priority < 0.3: low-priority (skip unless unique value)
+
+## Input Format
+[Chunk N] Priority: {priority}
+Source: [{filename}](./{filename}#{startLine}-{endLine})
+{content}
+Graph Nodes:
+  Node: {name} ({type}) | {properties}
+    → [{relation}] → target={targetType}:{targetName}
+
+## Output Format
+
+### Content Summary (REQUIRED)
+**Length: 1-3 sentences maximum**
+- Lead with highest-priority chunks ONLY
+- Skip chunks Priority < 0.3
+- Include lower-priority chunks only if they add unique value
+- Cite sources: [{filename}](./{filename}#{startLine})
+
+### Entity & Relation Map (CONDITIONAL)
+Only include if:
+1. Graph nodes are present AND
+2. They add meaningful connections not in Content Summary
+
+Format:
+**{EntityName}** ({type})
+  → [{relation}] → {TargetEntity} ({targetType})
+
+### Key Insights (CONDITIONAL)
+Only include if:
+1. Notable patterns exist AND
+2. They are NOT already covered in Content Summary
+
+Format:
+- 1-2 bullets maximum
+
+## Fixed Response
+If no chunks have Priority > 0.3 or all chunks are irrelevant:
+```
+No relevant knowledge found.
+```
+**DO NOT generate any other content.**
+
+## Examples
+
+### Good Output (relevant chunks)
+```
+The codebase uses React for frontend (Priority: 0.78).
+Source: [frontend/index.tsx](./frontend/index.tsx#L10-L50)
+```
+
+### Bad Output (NEVER generate)
+```
+Hello! Based on the documents, I can help you with React setup.
+The project structure follows a standard pattern.
+You should use TypeScript for type safety.
+```
+(These weren't in the input!)
+
+## Strict Rules
+- Output ONLY content from input chunks
+- Skip all sections with no data
+- Maximum 3 sentences in Content Summary
+- NEVER exceed provided information
+- Cite sources when including facts
+- If uncertain: "No relevant knowledge found."

package/dist/assets/txt/model/memory-summary.txt

@@ -1,40 +1,80 @@
-# Memory Summary Prompt
-
-## Task
-Synthesize the provided memory facts into ONE concise, structured summary relevant to the query.
-Each fact has a pre-computed **Priority** score (0-1). Use it as the sole ordering signal.
-
-## Priority Score
-Priority = 0.35 × relevance + 0.35 × recency + 0.30 × importance_normalized
-
-Higher Priority = more relevant, more recent, and more important.
-Facts with Priority > 0.6 are high-priority. Facts < 0.3 are low-priority.
-When a high-priority fact contradicts a low-priority one, the high-priority fact is the current truth.
-
-## Input Format
-[Fact N] Priority: {priority}
-Category: {category} | Importance: {importance} | Source: {source}
-{content}
-Graph Nodes:
-  Node: {name} ({type})
-    --[{relation}]--> {targetName} ({targetType})
-
-## Output
-
-### Content Summary
-2-4 sentences. Lead with highest-priority facts.
-Include lower-priority facts only if they add unique value not covered by higher-priority ones.
-If a high-priority fact updates or contradicts a lower-priority one, state the current truth clearly.
-
-### Entity & Relation Map (only if graph nodes are present)
-Deduplicate nodes by name, merge relations across all facts:
-**{EntityName}** ({type})
-  → [{relation}] → {TargetEntity} ({targetType})
-
-### Key Insights (only if meaningful patterns exist)
-1-2 bullets. Highlight contradictions between high and low priority facts.
-
-## Rules
-- Skip sections with no data
-- Base output only on provided content, do not invent facts
-- If no relevant information: "No relevant memory found."
+# Memory Summary Prompt
+
+## Task
+Synthesize memory facts into a concise summary relevant to the query.
+
+**CRITICAL RULES:**
+- NEVER generate facts not present in the input
+- NEVER invent content or relationships
+- NEVER elaborate beyond provided information
+- NEVER guess or assume user intent
+- If input has no valuable information: ONLY output "No relevant memory found."
+
+## Priority Score
+Priority = 0.4 × relevance + 0.3 × recency + 0.3 × importance
+- Priority > 0.6: high-priority (include)
+- Priority < 0.3: low-priority (skip unless unique value)
+
+## Input Format
+[Fact N] Priority: {priority}
+Category: {category} | Importance: {importance} | Source: {source}
+{content}
+Graph Nodes:
+  Node: {name} ({type})
+    → [{relation}] → target={targetType}:{targetName}
+
+## Output Format
+
+### Content Summary (REQUIRED)
+**Length: 1-3 sentences maximum**
+- Lead with highest-priority facts ONLY
+- Skip facts Priority < 0.3
+- Include low-priority facts only if they add unique value not in high-priority facts
+- State contradictions clearly: "High-priority fact X contradicts low-priority fact Y. Current truth: X."
+
+### Entity & Relation Map (CONDITIONAL)
+Only include if:
+1. Graph nodes are present AND
+2. They add meaningful connections not in Content Summary
+
+Format:
+**{EntityName}** ({type})
+  → [{relation}] → {TargetEntity} ({targetType})
+
+### Key Insights (CONDITIONAL)
+Only include if:
+1. Meaningful patterns exist AND
+2. They are NOT already covered in Content Summary
+
+Format:
+- 1-2 bullets maximum
+
+## Fixed Response
+If no facts have Priority > 0.3 or all facts are irrelevant:
+```
+No relevant memory found.
+```
+**DO NOT generate any other content.**
+
+## Examples
+
+### Good Output (relevant facts)
+```
+The user prefers dark mode (Priority: 0.72).
+High-priority fact contradicts earlier light mode preference.
+```
+
+### Bad Output (NEVER generate)
+```
+Hello! I can help you with dark mode settings.
+The user's name is John and they work at Company X.
+They like TypeScript and use VSCode daily.
+```
+(These facts weren't in the input!)
+
+## Strict Rules
+- Output ONLY facts from input
+- Skip all sections with no data
+- Maximum 3 sentences in Content Summary
+- NEVER exceed provided information
+- If uncertain: "No relevant memory found."