contextsynapse 1.0.0__py3-none-any.whl

contextsynapse/config/schemas/api_spec.yaml

@@ -0,0 +1,48 @@
+# API Specification Extraction Schema
+# Used when ingesting OpenAPI specs, API docs, REST/GraphQL references, and SDK docs.
+
+name: api_spec
+
+node_types:
+  Endpoint:
+    fields: [path, method, summary, description, tags, deprecated, version]
+    required: [path, method]
+    description: "An API endpoint or operation"
+
+  Parameter:
+    fields: [name, location, type, required, description, default, example]
+    required: [name]
+    description: "A request parameter (query, path, header, or body field)"
+
+  Response:
+    fields: [status_code, description, content_type, schema]
+    required: [status_code]
+    description: "A response definition for an API endpoint"
+
+  ErrorCode:
+    fields: [code, message, description, resolution]
+    required: [code]
+    description: "An API error code and its meaning"
+
+edge_types:
+  ACCEPTS:
+    source: Endpoint
+    target: Parameter
+  RETURNS:
+    source: Endpoint
+    target: Response
+  RAISES:
+    source: Endpoint
+    target: ErrorCode
+  CALLS:
+    source: Endpoint
+    target: Endpoint
+
+fact_types:
+  - type: RateLimit
+    fields: [limit, period, scope, source_text]
+    description: "An API rate limit or throttling rule"
+
+  - type: AuthRequirement
+    fields: [type, description, scopes, source_text]
+    description: "An authentication or authorization requirement for an endpoint"

contextsynapse/config/schemas/architecture_doc.yaml

@@ -0,0 +1,51 @@
+# Architecture Document Extraction Schema
+# Used when ingesting system design docs, ADRs, infrastructure diagrams, and runbooks.
+
+name: architecture_doc
+
+node_types:
+  Component:
+    fields: [name, type, language, team, version, description]
+    required: [name]
+    description: "A software component, module, or library"
+
+  Service:
+    fields: [name, type, protocol, port, host, description, sla]
+    required: [name]
+    description: "A running service, microservice, or API"
+
+  Database:
+    fields: [name, type, engine, version, host, description]
+    required: [name]
+    description: "A database or persistent storage system"
+
+  Deployment:
+    fields: [name, environment, platform, region, replicas, description]
+    required: [name]
+    description: "A deployment target or environment (prod, staging, cloud region)"
+
+edge_types:
+  DEPENDS_ON:
+    source: Component
+    target: Component
+  CONNECTS_TO:
+    source: Service
+    target: Service
+  DEPLOYED_ON:
+    source: Service
+    target: Deployment
+  READS_FROM:
+    source: Service
+    target: Database
+  WRITES_TO:
+    source: Service
+    target: Database
+
+fact_types:
+  - type: DesignDecision
+    fields: [title, rationale, alternatives, consequences, source_text]
+    description: "An architectural decision or design choice documented"
+
+  - type: NonFunctionalRequirement
+    fields: [name, category, value, unit, source_text]
+    description: "A non-functional requirement such as latency, throughput, or availability"

contextsynapse/config/schemas/chat_history.yaml

@@ -0,0 +1,90 @@
+# Chat History Extraction Schema
+# Used when ingesting AI chat exports (ChatGPT, Claude, Gemini).
+# The pipeline parses conversations into text, then extracts these node/edge types.
+
+name: chat_history
+
+node_types:
+  Conversation:
+    fields: [name, source, model, message_count, created_at, updated_at]
+    required: [name]
+    description: "A single chat conversation/thread"
+
+  Topic:
+    fields: [name, description, category]
+    required: [name]
+    description: "A topic or subject discussed in the conversation"
+
+  Person:
+    fields: [name, role, affiliation]
+    required: [name]
+    description: "A person mentioned in the conversation"
+
+  Organization:
+    fields: [name, type, industry]
+    required: [name]
+    description: "A company, product, or organization discussed"
+
+  Concept:
+    fields: [name, description, domain]
+    required: [name]
+    description: "A technical concept, framework, or idea discussed"
+
+  Decision:
+    fields: [statement, rationale, context]
+    required: [statement]
+    description: "A decision or conclusion reached during the conversation"
+
+  Task:
+    fields: [title, description, status, assigned_to]
+    required: [title]
+    description: "A task, action item, or todo discussed or created"
+
+  Fact:
+    fields: [statement, source, confidence]
+    required: [statement]
+    description: "A factual claim, instruction, or preference expressed"
+
+  CodeSnippet:
+    fields: [name, language, description, purpose]
+    required: [name]
+    description: "A code snippet, function, or technical solution discussed"
+
+edge_types:
+  DISCUSSES:
+    source: Conversation
+    target: Topic
+  MENTIONS:
+    source: Conversation
+    target: Person
+  REFERENCES:
+    source: Conversation
+    target: Organization
+  USES_CONCEPT:
+    source: Conversation
+    target: Concept
+  DECIDED:
+    source: Conversation
+    target: Decision
+  ASSIGNED:
+    source: Conversation
+    target: Task
+  RELATED_TO:
+    source: Topic
+    target: Topic
+  WORKS_FOR:
+    source: Person
+    target: Organization
+
+fact_types:
+  - type: Preference
+    fields: [statement, confidence, context]
+    description: "A user preference or instruction expressed across conversations"
+
+  - type: Insight
+    fields: [statement, source_conversation, confidence]
+    description: "A key insight or learning from the conversation"
+
+  - type: Instruction
+    fields: [statement, context, scope]
+    description: "An instruction, rule, or guideline discussed"

contextsynapse/config/schemas/chatgpt_conversation.yaml

@@ -0,0 +1,101 @@
+# ChatGPT Conversation Extraction Schema
+# Optimized for ChatGPT data exports (conversations.json from Settings → Export).
+# ChatGPT conversations tend to be longer, more code-heavy, and multi-turn.
+
+name: chatgpt_conversation
+
+strategy_hints: |
+  This text is from a ChatGPT conversation between a user and an assistant.
+  The user asks questions and the assistant provides detailed solutions.
+  Focus on extracting:
+  - TOOLS and TECHNOLOGIES mentioned (Python, React, AWS, Docker, PostgreSQL, etc.)
+  - CONCEPTS discussed (machine learning, API design, data modeling, etc.)
+  - PROBLEMS the user faced (errors, bugs, design challenges, questions)
+  - SOLUTIONS the assistant provided (code, commands, architecture, explanations)
+  - DECISIONS made during the conversation (chose X approach because Y)
+  - CODE patterns and libraries used (specific functions, packages, commands)
+  Extract at least 3-5 entities per conversation. Every tool, library, error message, and technical term counts.
+
+node_types:
+  Topic:
+    fields: [name, description, category]
+    required: [name]
+    description: "A topic or subject discussed (e.g., 'database optimization', 'React hooks', 'deployment')"
+
+  Concept:
+    fields: [name, description, domain, complexity]
+    required: [name]
+    description: "A technical concept, pattern, or methodology (e.g., 'REST API', 'dependency injection', 'backpropagation')"
+
+  Tool:
+    fields: [name, type, purpose, version]
+    required: [name]
+    description: "A software tool, library, framework, or service (e.g., 'pandas', 'Next.js', 'Redis', 'GPT-4')"
+
+  Decision:
+    fields: [statement, rationale, alternatives_considered]
+    required: [statement]
+    description: "A design decision, architecture choice, or recommendation"
+
+  Problem:
+    fields: [statement, context, error_message]
+    required: [statement]
+    description: "A problem, error, bug, or challenge the user faced"
+
+  Solution:
+    fields: [statement, approach, tools_used, code_language]
+    required: [statement]
+    description: "A solution, fix, code snippet, or recommendation provided"
+
+  Task:
+    fields: [title, description, status]
+    required: [title]
+    description: "An action item or next step discussed"
+
+  Person:
+    fields: [name, role, organization]
+    required: [name]
+    description: "A real person mentioned (not the user or ChatGPT)"
+
+  Organization:
+    fields: [name, type, industry]
+    required: [name]
+    description: "A company, product, or platform mentioned"
+
+  Fact:
+    fields: [statement, source, confidence]
+    required: [statement]
+    description: "A factual claim, configuration detail, or technical specification"
+
+edge_types:
+  DISCUSSES:
+    source: Topic
+    target: Concept
+  SOLVED_BY:
+    source: Problem
+    target: Solution
+  USES:
+    source: Solution
+    target: Tool
+  DECIDED:
+    source: Decision
+    target: Concept
+  RELATED_TO:
+    source: Concept
+    target: Concept
+  IMPLEMENTS:
+    source: Tool
+    target: Concept
+
+fact_types:
+  - type: Configuration
+    fields: [statement, tool, context]
+    description: "A configuration, command, or technical setting"
+
+  - type: CodePattern
+    fields: [statement, language, library]
+    description: "A code pattern, function usage, or API call"
+
+  - type: BestPractice
+    fields: [statement, domain, confidence]
+    description: "A best practice or recommendation"

contextsynapse/config/schemas/claude_conversation.yaml

@@ -0,0 +1,148 @@
+# Claude Conversation Extraction Schema
+# Optimized for claude.ai conversation exports.
+# Extracts topics, technical concepts, decisions, tasks, and solutions.
+
+name: claude_conversation
+
+extraction:
+  strategy: conversation
+  parser: turn
+  significance_threshold: 0.3
+  signals: [decision, action, question, problem, solution]
+  topic_clustering: true
+  topic_method: llm_with_fallback
+  stages:
+    - detect_signals
+    - extract_entities
+    - store_documents
+    - resolve_entities
+    - build_edges
+    - infer_domains
+    - cluster_topics
+    - link_cross_reference
+    - validate_gate
+    - deduplicate
+    - synthesize_cu
+    - embed
+    - index_bm25
+
+strategy_hints: |
+  This text is from an AI chat conversation between a user and an assistant.
+  The user asks questions and the assistant provides solutions.
+  Focus on extracting:
+  - TOOLS and TECHNOLOGIES mentioned (AWS, Docker, React, Blender, etc.)
+  - CONCEPTS discussed (reinforcement learning, system design, etc.)
+  - PROBLEMS the user faced (errors, bugs, design challenges)
+  - SOLUTIONS the assistant provided (commands, code, architecture recommendations)
+  - DECISIONS: only extract if someone explicitly chose X over Y with reasoning. NOT random statements or instructions.
+  - TASKS or action items discussed
+
+  IMPORTANT RULES:
+  - A Decision MUST be an explicit choice between alternatives with rationale. NOT random statements.
+  - An Action MUST be a specific task someone will do. NOT modal verbs like "must", "should", "could".
+  - A Person MUST be a real named human (e.g. "Elon Musk", "Dr. Sharma"). NOT concepts, roles, or agent names.
+  - Do NOT extract single words as entities. Every entity name must be 2+ words or a proper noun.
+  - Do NOT extract UI descriptions, code output, formatting instructions, or code variable names.
+  - Only extract entities that are meaningful, specific, and reusable. Skip noise aggressively.
+
+node_types:
+  Topic:
+    fields: [name, description, category]
+    required: [name]
+    description: "A topic or subject discussed (e.g., 'SSH troubleshooting', 'meal planning', 'system design')"
+
+  Concept:
+    fields: [name, description, domain, complexity]
+    required: [name]
+    description: "A technical concept, framework, technology, or methodology discussed (e.g., 'React', 'Kubernetes', 'reinforcement learning')"
+
+  Tool:
+    fields: [name, type, purpose, url]
+    required: [name]
+    description: "A software tool, service, library, or platform mentioned (e.g., 'AWS EC2', 'Blender', 'AutoGen', 'Docker')"
+
+  Decision:
+    fields: [statement, rationale, alternatives_considered]
+    required: [statement]
+    description: "A design decision, architecture choice, or recommendation made during the conversation"
+
+  Problem:
+    fields: [statement, context, severity]
+    required: [statement]
+    description: "A problem, error, bug, or challenge discussed"
+
+  Solution:
+    fields: [statement, approach, tools_used]
+    required: [statement]
+    description: "A solution, fix, workaround, or recommendation provided"
+
+  Task:
+    fields: [title, description, status, priority]
+    required: [title]
+    description: "An action item, next step, or task discussed"
+
+  Person:
+    fields: [name, role, organization]
+    required: [name]
+    description: "A real person mentioned (not the user or assistant)"
+
+  Organization:
+    fields: [name, type, industry]
+    required: [name]
+    description: "A company, team, or organization mentioned"
+
+  Fact:
+    fields: [statement, source, confidence]
+    required: [statement]
+    description: "A factual claim, configuration detail, or technical specification"
+
+edge_types:
+  DISCUSSES:
+    source: Topic
+    target: Concept
+  SOLVED_BY:
+    source: Problem
+    target: Solution
+  USES:
+    source: Solution
+    target: Tool
+  DECIDED:
+    source: Decision
+    target: Concept
+  RELATED_TO:
+    source: Concept
+    target: Concept
+  IMPLEMENTS:
+    source: Tool
+    target: Concept
+  USED_WITH:
+    source: Tool
+    target: Tool
+  ADDRESSES:
+    source: Solution
+    target: Problem
+  PART_OF:
+    source: Tool
+    target: Topic
+  INVOLVES:
+    source: Topic
+    target: Tool
+  ABOUT:
+    source: Problem
+    target: Concept
+  RECOMMENDS:
+    source: Decision
+    target: Tool
+
+fact_types:
+  - type: Configuration
+    fields: [statement, tool, context]
+    description: "A configuration setting, command, or technical detail"
+
+  - type: Recommendation
+    fields: [statement, confidence, context]
+    description: "A best practice or recommendation from the conversation"
+
+  - type: Learning
+    fields: [statement, source_conversation, domain]
+    description: "A key insight or learning from the discussion"

contextsynapse/config/schemas/contract.yaml

@@ -0,0 +1,48 @@
+# Contract Extraction Schema
+# Used when ingesting legal contracts, agreements, SOWs, and NDAs.
+
+name: contract
+
+node_types:
+  Party:
+    fields: [name, type, address, jurisdiction, signatory, role]
+    required: [name]
+    description: "A party to the contract (individual or organization)"
+
+  Term:
+    fields: [title, description, section, category, effective_date, expiry_date]
+    required: [title]
+    description: "A contractual term, clause, or condition"
+
+  Obligation:
+    fields: [description, party, deadline, type, status]
+    required: [description]
+    description: "A specific obligation or duty imposed on a party"
+
+  Penalty:
+    fields: [description, amount, currency, trigger, party]
+    required: [description]
+    description: "A penalty, fee, or consequence for breach of the contract"
+
+edge_types:
+  BOUND_BY:
+    source: Party
+    target: Term
+  SUBJECT_TO:
+    source: Party
+    target: Obligation
+  GOVERNS:
+    source: Term
+    target: Obligation
+  TRIGGERS:
+    source: Obligation
+    target: Penalty
+
+fact_types:
+  - type: ContractDate
+    fields: [event, date, description, source_text]
+    description: "A significant date in the contract lifecycle (signing, effective, expiry)"
+
+  - type: FinancialTerm
+    fields: [description, amount, currency, schedule, conditions, source_text]
+    description: "A financial term, payment obligation, or compensation clause"

contextsynapse/config/schemas/conversation.yaml

@@ -0,0 +1,167 @@
+# Universal Conversation Extraction Schema
+# Optimized for any AI conversation export (ChatGPT, Claude, Gemini, etc.)
+# Extracts topics, technical concepts, decisions, tasks, solutions, interests, and preferences.
+
+name: conversation
+
+extraction:
+  strategy: conversation
+  parser: turn
+  significance_threshold: 0.3
+  signals: [decision, action, question, problem, solution]
+  topic_clustering: true
+  topic_method: llm_with_fallback
+  stages:
+    - detect_signals
+    - extract_entities
+    - extract_preferences
+    - store_documents
+    - resolve_entities
+    - build_edges
+    - infer_domains
+    - cluster_topics
+    - link_cross_reference
+    - validate_gate
+    - deduplicate
+    - synthesize_cu
+    - embed
+    - index_bm25
+
+strategy_hints: |
+  This text is from a conversation between a user and an AI assistant.
+  It could be from ChatGPT, Claude, Gemini, or any other AI.
+  Focus on extracting:
+  - TOOLS and TECHNOLOGIES mentioned (AWS, Docker, React, Python, etc.)
+  - CONCEPTS discussed (machine learning, insurance, statistics, etc.)
+  - PROBLEMS the user faced (errors, bugs, design challenges)
+  - SOLUTIONS the assistant provided (commands, code, recommendations)
+  - DECISIONS: only extract if someone explicitly chose X over Y with reasoning
+  - TASKS or action items discussed
+  - USER INTERESTS: topics the user repeatedly asks about or shows curiosity in
+
+  IMPORTANT RULES:
+  - A Decision MUST be an explicit choice between alternatives with rationale. NOT random statements.
+  - An Action MUST be a specific task someone will do. NOT modal verbs like "must", "should", "could".
+  - A Person MUST be a real named human (e.g. "Elon Musk", "Dr. Sharma"). NOT concepts, roles, or agent names like "Data Validation Agent" or "Functional Requirements".
+  - Do NOT extract single words as entities. Every entity name must be 2+ words or a proper noun.
+  - Do NOT extract UI descriptions, code output, formatting instructions, or code variable names.
+  - Do NOT extract generic phrases like "Updated Flow", "Chosen", "Failed" as entities.
+  - Only extract entities that are meaningful, specific, and reusable. Skip noise aggressively.
+
+node_types:
+  Topic:
+    fields: [name, description, category]
+    required: [name]
+    description: "A topic or subject discussed (e.g., 'SSH troubleshooting', 'meal planning', 'system design')"
+
+  Concept:
+    fields: [name, description, domain, complexity]
+    required: [name]
+    description: "A technical concept, framework, technology, or methodology discussed (e.g., 'React', 'Kubernetes', 'reinforcement learning')"
+
+  Tool:
+    fields: [name, type, purpose, url]
+    required: [name]
+    description: "A software tool, service, library, or platform mentioned (e.g., 'AWS EC2', 'Blender', 'AutoGen', 'Docker')"
+
+  Decision:
+    fields: [statement, rationale, alternatives_considered]
+    required: [statement]
+    description: "A design decision, architecture choice, or recommendation made during the conversation"
+
+  Problem:
+    fields: [statement, context, severity]
+    required: [statement]
+    description: "A problem, error, bug, or challenge discussed"
+
+  Solution:
+    fields: [statement, approach, tools_used]
+    required: [statement]
+    description: "A solution, fix, workaround, or recommendation provided"
+
+  Task:
+    fields: [title, description, status, priority]
+    required: [title]
+    description: "An action item, next step, or task discussed"
+
+  Person:
+    fields: [name, role, organization]
+    required: [name]
+    description: "A real named human mentioned (e.g. 'Elon Musk', 'Dr. Priya Sharma'). NOT roles, concepts, agent names, or software components."
+
+  Organization:
+    fields: [name, type, industry]
+    required: [name]
+    description: "A company, team, or organization mentioned"
+
+  Fact:
+    fields: [statement, source, confidence]
+    required: [statement]
+    description: "A factual claim, configuration detail, or technical specification"
+
+  Interest:
+    fields: [name, description, frequency]
+    required: [name]
+    description: "A topic or area the user shows repeated interest in across conversations"
+
+  Preference:
+    fields: [statement, category, confidence]
+    required: [statement]
+    description: "A user preference or working style detected from conversation patterns"
+
+edge_types:
+  DISCUSSES:
+    source: Topic
+    target: Concept
+  SOLVED_BY:
+    source: Problem
+    target: Solution
+  USES:
+    source: Solution
+    target: Tool
+  DECIDED:
+    source: Decision
+    target: Concept
+  RELATED_TO:
+    source: Concept
+    target: Concept
+  IMPLEMENTS:
+    source: Tool
+    target: Concept
+  USED_WITH:
+    source: Tool
+    target: Tool
+  ADDRESSES:
+    source: Solution
+    target: Problem
+  PART_OF:
+    source: Tool
+    target: Topic
+  INVOLVES:
+    source: Topic
+    target: Tool
+  ABOUT:
+    source: Problem
+    target: Concept
+  RECOMMENDS:
+    source: Decision
+    target: Tool
+  INTERESTED_IN:
+    source: Session
+    target: Interest
+  PREFERS:
+    source: Session
+    target: Preference
+
+fact_types:
+  - type: Configuration
+    fields: [statement, tool, context]
+    description: "A configuration setting, command, or technical detail"
+
+  - type: Recommendation
+    fields: [statement, confidence, context]
+    description: "A best practice or recommendation from the conversation"
+
+  - type: Learning
+    fields: [statement, source_conversation, domain]
+    description: "A key insight or learning from the discussion"