Skip to main content

​
Overview

Connect skills with Upsonic’s KnowledgeBase to enable semantic search across all skill reference documents. Pass the KnowledgeBase instance via context=[kb] on the Task so the agent can query reference documents at runtime.

​
Usage

from upsonic import Agent, Task, KnowledgeBase
from upsonic.skills import Skills, LocalSkills
from upsonic.embeddings import GeminiEmbedding, GeminiEmbeddingConfig
from upsonic.vectordb import ChromaProvider, ChromaConfig, ConnectionConfig, Mode

# Setup embedding provider (uses GOOGLE_API_KEY env var)
embedding = GeminiEmbedding(GeminiEmbeddingConfig())

# Setup vector database (embedded mode — no external server needed)
config = ChromaConfig(
    collection_name="skill_references",
    vector_size=3072,  # Gemini default dimension
    connection=ConnectionConfig(mode=Mode.EMBEDDED, db_path="./chroma_db"),
)
vectordb = ChromaProvider(config)

# Create knowledge base from skill reference documents
kb = KnowledgeBase(
    sources=["./my-skills"],
    embedding_provider=embedding,
    vectordb=vectordb,
)

# Create skills
skills = Skills(
    loaders=[LocalSkills("./my-skills")],
)

agent = Agent(
    model="anthropic/claude-sonnet-4-6",
    name="Research Agent",
    role="Research Specialist",
    goal="Find and synthesize information from skill references",
    skills=skills,
)

# Pass the knowledge base via context on the Task
task = Task(
    description="Find best practices for error handling across all our skill references.",
    context=[kb],
)

result = agent.do(task)
print(result)

​
Using KnowledgeBase as a Tool

You can also pass the KnowledgeBase via tools=[kb] on the Task. This registers the KB as a searchable tool that the agent can actively query during execution: 
from upsonic import Agent, Task, KnowledgeBase
from upsonic.skills import Skills, LocalSkills
from upsonic.embeddings import GeminiEmbedding, GeminiEmbeddingConfig
from upsonic.vectordb import ChromaProvider, ChromaConfig, ConnectionConfig, Mode

embedding = GeminiEmbedding(GeminiEmbeddingConfig())
config = ChromaConfig(
    collection_name="skill_references",
    vector_size=3072,
    connection=ConnectionConfig(mode=Mode.EMBEDDED, db_path="./chroma_db"),
)
vectordb = ChromaProvider(config)

kb = KnowledgeBase(
    sources=["./my-skills"],
    embedding_provider=embedding,
    vectordb=vectordb,
)

skills = Skills(
    loaders=[LocalSkills("./my-skills")],
)

agent = Agent(
    model="anthropic/claude-sonnet-4-6",
    name="Research Agent",
    role="Research Specialist",
    goal="Find and synthesize information from skill references",
    skills=skills,
)

# Pass the knowledge base as a tool — agent can search it on demand
task = Task(
    description="Find best practices for error handling across all our skill references.",
    tools=[kb],
)

result = agent.do(task)
print(result)

​
How It Works

  • context=[kb] — The KB content is injected into the task context before execution. The agent receives it as background information.
  • tools=[kb] — The KB is registered as a callable tool. The agent can actively search it during execution to find relevant documents on demand.