qtype 0.1.11__py3-none-any.whl → 0.1.13__py3-none-any.whl

qtype/examples/legacy/rag.qtype.yaml

@@ -0,0 +1,207 @@
+id: rag_example
+description: |
+  End-to-end RAG (Retrieval Augmented Generation) example that processes documents 
+  into a Qdrant vector store and enables conversational chat with the indexed content.
+  
+  This example includes two flows:
+  
+  1. rag_chat: Chat with your documents using RAG
+     - Accept user questions as ChatMessage
+     - Extract text content from chat message using FieldExtractor
+     - Search vector index for relevant context using VectorSearch
+     - Generate contextual responses using Amazon Nova Lite
+     - Maintain conversation history with memory
+  
+  2. document_ingestion: Ingest documents into the vector store
+     - Load LlamaIndex Q&A pairs from HuggingFace (1235 instruction-output pairs)
+     - Uses AlignmentLab-AI/llama-index dataset via HuggingFaceFSReader
+     - Split documents into manageable chunks
+     - Generate embeddings using AWS Bedrock Titan Embed Text v2
+     - Store embedded chunks in Qdrant vector database using IndexUpsert
+  
+  Prerequisites:
+  - AWS credentials configured (via AWS CLI profile or environment variables)
+  - Qdrant running locally on port 6333 (or update the index args for Qdrant Cloud)
+  - Python dependencies: llama-index-readers-huggingface-fs (install with: uv add llama-index-readers-huggingface-fs --optional interpreter)
+  
+  To run ingestion:
+    uv run python -m qtype.cli run examples/rag.qtype.yaml --flow document_ingestion
+  
+  To start chat:
+    uv run python -m qtype.cli run examples/rag.qtype.yaml --flow rag_chat
+
+# AWS Authentication for Bedrock
+auths:
+  - type: aws
+    id: aws_auth
+    profile_name: ${AWS_PROFILE}  # Uses AWS_PROFILE env var
+
+# Models
+models:
+  # Embedding model using AWS Bedrock Titan v2
+  - type: EmbeddingModel
+    id: titan_embed_v2
+    provider: aws-bedrock
+    model_id: amazon.titan-embed-text-v2:0
+    dimensions: 1024
+    auth: aws_auth
+  
+  # Generative model for chat
+  - type: Model
+    id: claude_sonnet
+    provider: aws-bedrock
+    model_id: amazon.nova-lite-v1:0
+    inference_params:
+      temperature: 0.7
+      max_tokens: 2048
+    auth: aws_auth
+
+# Qdrant vector index for storing embedded documents
+indexes:
+  - type: VectorIndex
+    module: llama_index.vector_stores.qdrant.QdrantVectorStore
+    id: rag_index
+    name: documents
+    embedding_model: titan_embed_v2
+    args:
+      # Qdrant vector store arguments.
+      # See https://developers.llamaindex.ai/python/framework-api-reference/storage/vector_store/qdrant/
+      collection_name: documents
+      url: http://localhost:6333
+      api_key: ""  # Empty string for local Qdrant (library validation bug requires this)
+
+# Flow to ingest documents into the vector store
+flows:
+  # Flow for conversational RAG
+  - type: Flow
+    id: rag_chat
+    description: Chat with the document collection using RAG
+    
+    interface:
+      type: Conversational
+    
+    variables:
+      - id: user_message
+        type: ChatMessage
+      - id: user_question
+        type: text
+      - id: search_results
+        type: list[RAGSearchResult]
+      - id: context_prompt
+        type: text
+      - id: assistant_response
+        type: ChatMessage
+    
+    inputs:
+      - user_message
+    
+    outputs:
+      - assistant_response
+    
+    steps:
+      # Step 1: Extract text content from chat message blocks where type is text
+      - id: extract_question
+        type: FieldExtractor
+        json_path: "$.blocks[?(@.type == 'text')].content"
+        inputs:
+          - user_message
+        outputs:
+          - user_question
+      
+      # Step 2: Search the vector index for relevant chunks (VectorSearch handles embedding internally)
+      - id: search_index
+        type: VectorSearch
+        index: rag_index
+        default_top_k: 5
+        inputs:
+          - user_question
+        outputs:
+          - search_results
+      
+      # Step 3: Build context prompt with retrieved chunks
+      - id: build_prompt
+        type: PromptTemplate
+        template: |
+          You are a helpful assistant that answers questions based on the provided context.
+          
+          Context from documents:
+          {search_results}
+          
+          User question: {user_question}
+          
+          Please provide a detailed answer based on the context above. If the context doesn't contain relevant information, say so.
+        inputs:
+          - search_results
+          - user_question
+        outputs:
+          - context_prompt
+      
+      # Step 4: Generate response using LLM with context
+      - id: generate_response
+        type: LLMInference
+        model: claude_sonnet
+        system_message: "You are a helpful assistant that answers questions based on provided document context. Be concise and accurate."
+        inputs:
+          - context_prompt
+        outputs:
+          - assistant_response
+
+  - type: Flow
+    id: document_ingestion
+    description: Load LlamaIndex Q&A pairs from HuggingFace, split, embed, and index documents
+    
+    variables:
+      - id: raw_document
+        type: RAGDocument
+      - id: document_chunk
+        type: RAGChunk
+      - id: embedded_chunk
+        type: RAGChunk
+    
+    outputs:
+      - embedded_chunk
+    
+    steps:
+      # Step 1: Load documents directly from HuggingFace using HuggingFaceFSReader
+      # This dataset contains 1235 LlamaIndex instruction-output Q&A pairs
+      - id: load_documents
+        type: DocumentSource
+        reader_module: llama_index.readers.huggingface_fs.HuggingFaceFSReader
+        loader_args:
+          path: "datasets/AlignmentLab-AI/llama-index/modified_dataset.jsonl"
+        outputs:
+          - raw_document
+      
+      # Step 2: Split documents into chunks
+      - id: split_documents
+        type: DocumentSplitter
+        splitter_name: "SentenceSplitter"
+        chunk_size: 512
+        chunk_overlap: 50
+        inputs:
+          - raw_document
+        outputs:
+          - document_chunk
+      
+      # Step 3: Generate embeddings for each chunk
+      - id: embed_chunks
+        type: DocumentEmbedder
+        model: titan_embed_v2
+        concurrency_config:
+          num_workers: 5
+        inputs:
+          - document_chunk
+        outputs:
+          - embedded_chunk
+      
+      # Step 4: Upsert embedded chunks into Qdrant
+      - id: index_chunks
+        type: IndexUpsert
+        index: rag_index
+        batch_config:
+          batch_size: 25
+        inputs:
+          - embedded_chunk
+        outputs:
+          - embedded_chunk
+

qtype/examples/legacy/sample_data.txt

@@ -0,0 +1,43 @@
+Sales Data Q1 2024
+==================
+
+Summary Statistics:
+- Total Records: 1,472 transactions
+- Date Range: January 1 - March 31, 2024
+- Products: A, B, C, D
+- Regions: North, South, East, West
+
+Top Performing Products:
+1. Product A: $1,245,890 total sales (avg $847/transaction)
+2. Product C: $1,123,450 total sales (avg $763/transaction)
+3. Product B: $987,230 total sales (avg $671/transaction)
+4. Product D: $856,340 total sales (avg $582/transaction)
+
+Regional Breakdown:
+- North Region: $1,089,450 (26.3% of total)
+- South Region: $1,045,230 (25.2% of total)
+- East Region: $1,012,890 (24.4% of total)
+- West Region: $1,001,340 (24.1% of total)
+
+Key Trends:
+- Product A shows consistent growth throughout Q1 (+15% month-over-month)
+- West region had strongest growth rate despite lower absolute numbers
+- Discount rates ranged from 0% to 20%, with 10% being most common
+- Average transaction size increased 8% from Jan to March
+- Customer count per transaction averaging 25 customers
+
+Notable Patterns:
+- Weekday sales 23% higher than weekend sales
+- Product D underperforming in South region specifically
+- Strong correlation between discount rate and quantity sold
+- North region showing premium pricing power (higher avg transaction)
+
+Concerns:
+- Product D sales declining 5% month-over-month
+- South region discount rates higher than other regions
+- Customer acquisition costs need monitoring
+
+Opportunities:
+- West region expansion showing strong ROI
+- Product A momentum could be leveraged for upselling
+- Cross-regional best practices could improve South region performance

qtype/examples/legacy/time_utilities.qtype.yaml

@@ -0,0 +1,64 @@
+id: time_utilities
+description: |
+  A simple application demonstrating tool usage with the commons library.
+  
+  This example shows how to:
+  - Import tools from the commons library
+  - Invoke Python functions as tools
+  - Pass data between tool invocations
+  - Work with datetime types
+  
+  The flow gets the current timestamp, formats it, calculates a future time,
+  and determines the time difference.
+
+# Import the commons tools library
+references:
+  - !include ../common/tools.qtype.yaml
+
+flows:
+  - type: Flow
+    id: time_info_flow
+    description: Get and format the current timestamp
+    
+    variables:
+      - id: current_time
+        type: datetime
+      
+      - id: time_two_hours_later
+        type: datetime
+      
+      - id: time_difference
+        type: TimeDifferenceResultType
+    
+    outputs:
+      - current_time
+      - time_difference
+    
+    steps:
+      # Step 1: Get current timestamp
+      - id: get_time
+        type: InvokeTool
+        tool: qtype.application.commons.tools.get_current_timestamp
+        input_bindings: {}
+        output_bindings:
+          result: current_time
+      
+      # Step 2: Calculate time 2 hours from now
+      - id: add_hours
+        type: InvokeTool
+        tool: qtype.application.commons.tools.timedelta
+        input_bindings:
+          timestamp: current_time
+          hours: "2"
+        output_bindings:
+          result: time_two_hours_later
+      
+      # Step 3: Calculate the time difference
+      - id: calc_difference
+        type: InvokeTool
+        tool: qtype.application.commons.tools.calculate_time_difference
+        input_bindings:
+          start_time: current_time
+          end_time: time_two_hours_later
+        output_bindings:
+          result: time_difference

qtype/examples/legacy/vertex/README.md

@@ -0,0 +1,11 @@
+To use a Vertex example, set the following environment variables manually:
+```
+export GOOGLE_CLOUD_PROJECT=
+export GOOGLE_CLOUD_LOCATION=
+export GOOGLE_APPLICATION_CREDENTIALS=
+```
+
+Using gcloud CLI:
+How to setup [Gemini CLI Authentication Setup](https://github.com/google-gemini/gemini-cli/blob/main/docs/get-started/authentication.md)
+
+`profile_name`: Replace default with relevant GCP project name in `.config/gcloud/configurations/your-project-configuration-file`

qtype/examples/legacy/vertex/hello_world_chat.qtype.yaml

@@ -0,0 +1,36 @@
+id: hello_world
+description: A simple stateful chat flow with Google Vertex AI
+models:
+  - type: Model
+    id: gemini_flash
+    provider: gcp-vertex
+    model_id: gemini-2.0-flash
+    inference_params:
+      temperature: 0.7
+memories:
+  - id: chat_memory
+    token_limit: 10000
+flows:
+  - type: Flow
+    id: simple_chat_example
+    interface:
+      type: Conversational
+    variables:
+      - id: user_message
+        type: ChatMessage
+      - id: response_message
+        type: ChatMessage
+    inputs:
+      - user_message
+    outputs:
+      - response_message
+    steps:
+      - id: llm_inference_step
+        type: LLMInference
+        model: gemini_flash
+        system_message: "You are a helpful assistant."
+        memory: chat_memory
+        inputs:
+          - user_message
+        outputs:
+          - response_message

qtype/examples/legacy/vertex/hello_world_completion.qtype.yaml

@@ -0,0 +1,40 @@
+id: hello_world
+description: A simple hello world application using Google Vertex AI
+models:
+  - type: Model
+    id: gemini_flash
+    provider: gcp-vertex
+    model_id: gemini-2.0-flash
+    inference_params:
+      temperature: 0.7
+flows:
+  - type: Flow
+    id: simple_example
+    interface:
+      type: Complete
+    variables:
+      - id: prompt
+        type: text
+      - id: formatted_prompt
+        type: text
+      - id: answer
+        type: text
+    inputs:
+      - prompt
+    outputs:
+      - answer
+    steps:
+      - id: question_prompt
+        type: PromptTemplate
+        template: "You are a helpful assistant. Answer the following question:\n{prompt}\n"
+        inputs:
+          - prompt
+        outputs:
+          - formatted_prompt
+      - id: llm_inference_step
+        type: LLMInference
+        model: gemini_flash
+        inputs:
+          - formatted_prompt
+        outputs:
+          - answer

qtype/examples/legacy/vertex/hello_world_completion_with_auth.qtype.yaml

@@ -0,0 +1,45 @@
+id: hello_world
+description: A simple hello world application using Google Vertex AI
+models:
+  - type: Model
+    id: gemini_flash
+    provider: gcp-vertex
+    model_id: gemini-2.0-flash
+    inference_params:
+      temperature: 0.7
+    auth: vertex_auth
+auths:
+  - type: vertex
+    id: vertex_auth
+    profile_name: default
+flows:
+  - type: Flow
+    id: simple_example
+    interface:
+      type: Complete
+    variables:
+      - id: prompt
+        type: text
+      - id: answer
+        type: text
+      - id: formatted_prompt
+        type: text
+    inputs:
+      - prompt
+    outputs:
+      - answer
+    steps:
+      - id: question_prompt
+        type: PromptTemplate
+        template: "You are a helpful assistant. Answer the following question:\n{prompt}\n"
+        inputs:
+          - prompt
+        outputs:
+          - formatted_prompt
+      - id: llm_inference_step
+        type: LLMInference
+        model: gemini_flash
+        inputs:
+          - formatted_prompt
+        outputs:
+          - answer

qtype/examples/observability_debugging/trace_with_opentelemetry.qtype.yaml

@@ -0,0 +1,40 @@
+id: trace_example
+description: Example of tracing QType application calls with OpenTelemetry to Phoenix
+
+models:
+  - type: Model
+    id: nova
+    provider: aws-bedrock
+    model_id: amazon.nova-lite-v1:0
+    inference_params:
+      temperature: 0.7
+      max_tokens: 512
+
+flows:
+  - type: Flow
+    id: classify_text
+    interface:
+      type: Complete
+    variables:
+      - id: text
+        type: text
+      - id: response
+        type: text
+    inputs:
+      - text
+    outputs:
+      - response
+    steps:
+      - id: classify
+        type: LLMInference
+        model: nova
+        system_message: "Classify the following text as positive, negative, or neutral. Respond with only one word."
+        inputs:
+          - text
+        outputs:
+          - response
+
+telemetry:
+  id: phoenix_trace
+  provider: Phoenix
+  endpoint: http://localhost:6006/v1/traces

qtype/examples/research_assistant/research_assistant.qtype.yaml

@@ -0,0 +1,94 @@
+# Research Assistant Example
+#
+# Takes one input (topic), searches the web via Tavily, then synthesizes
+# an answer using an LLM call.
+#
+# Prereqs:
+# - Set `TAVILY-API_BEARER` for Tavily auth (see tavily.qtype.yaml)
+# - Configure AWS credentials if using Bedrock (default model below)
+#
+# Run with:
+#   qtype run \
+#     -i '{"topic":"Latest developments in retrieval augmented generation"}' \
+#     examples/research_assistant/research_assistant.qtype.yaml
+
+id: research_assistant
+description: Web search + synthesis research assistant using Tavily
+
+# Import Tavily tools created from tavily.oas.yaml
+references:
+  - !include ./tavily.qtype.yaml
+
+models:
+  - type: Model
+    id: nova_lite
+    provider: aws-bedrock
+    model_id: amazon.nova-lite-v1:0
+    inference_params:
+      temperature: 0.3
+      max_tokens: 900
+
+flows:
+  - type: Flow
+    id: research
+    description: Search the web for a topic and synthesize an answer
+    inputs:
+      - topic
+    outputs:
+      - answer
+
+    variables:
+      - id: topic
+        type: text
+
+      # Tavily outputs
+      - id: tavily_results
+        type:
+          element_type: schema_4844016144
+
+      # LLM prompt + response
+      - id: synthesis_prompt
+        type: text
+      - id: answer
+        type: text
+
+    steps:
+      - type: InvokeTool
+        id: search_web
+        tool: search
+        input_bindings:
+          query: topic
+        output_bindings:
+          results: tavily_results
+        outputs:
+          - tavily_results
+
+      - type: PromptTemplate
+        id: build_prompt
+        template: |
+          Research topic: {topic}
+
+          Search results (list of objects with url/content/score):
+          {tavily_results}
+
+          Task:
+          - Write a concise, well-structured answer to the research topic.
+          - If the results contain URLs, include 3-8 bullet citations at the end
+            using the URLs you relied on most.
+          - If information is missing or uncertain, say so explicitly.
+        inputs:
+          - topic
+          - tavily_results
+        outputs:
+          - synthesis_prompt
+
+      - type: LLMInference
+        id: synthesize
+        model: nova_lite
+        system_message: |
+          You are a careful research assistant. Use the provided search results.
+          Prefer accurate summaries over speculation.
+        inputs:
+          - synthesis_prompt
+        outputs:
+          - answer