Files
YiemAgent/learning/README.md
T
2026-07-31 11:41:53 +07:00

575 lines
17 KiB
Markdown

# AgentCore.jl - Learning Guide
## How to Use This Documentation
### Top-Down Learning Approach
This documentation is organized in a **top-down** order, starting from high-level concepts and drilling down into implementation details. Follow this sequence:
1. **Architecture Overview** - Understand the big picture
2. **Agent Component** - Learn about state management and event streaming
3. **AgentLoop Component** - Understand the core LLM interaction loop
4. **Types & Messages** - Learn the data structures
5. **Session Management** - Understand conversation history
6. **Tools** - Learn about tool execution
### Learning Style
- **Visual learners**: Study the ASCII diagrams
- **Hands-on learners**: Code examples provided for each section
- **Conceptual learners**: Read summaries and overviews first
## Quick Start
### Minimal Example
```julia
using AgentCore
# Create agent
agent = Agent(Dict(
:systemPrompt => "You are a helpful assistant.",
:model => Model(...),
:tools => [bash_tool],
))
# Run conversation
prompt(agent, "Hello!")
# Wait for completion
waitForIdle(agent)
```
### Understanding the Flow
```
User Code
├─► Create Agent
│ ├─ Initialize state
│ ├─ Set up queues
│ └─ Register hooks
├─► prompt("Hello")
│ ├─ Validate input
│ └─ Start AgentLoop
├─► AgentLoop (runs in thread)
│ ├─ Stream LLM response
│ ├─ Execute tools
│ └─ Emit events
└─► Event handlers receive events
├─ MessageEndEvent
├─ ToolExecutionEndEvent
└─ AgentEndEvent
```
## Core Concepts
### Agent
**What it is**: High-level interface for LLM interactions
**What it does**:
- Manages conversation state
- Handles event streaming
- Queues steering/follow-up messages
- Provides hooks for customization
**Key methods**:
- `prompt()` - Start new conversation
- `continue!()` - Continue existing conversation
- `steer()` - Queue message for next turn
- `followUp()` - Queue message after stop
- `subscribe()` - Listen to events
- `waitForIdle()` - Wait for agent to finish processing
- `reset!()` - Clear transcript state and queued messages
- `clearAllQueues()` - Remove all queued steering and follow-up messages
- `hasQueuedMessages()` - Check if queues have pending messages
- `abort()` - Abort the current run
- `get_state()` - Get the current agent state
### AgentLoop
**What it is**: Core LLM interaction loop
**What it does**:
- Calls LLM API with streaming
- Executes tool calls (parallel or sequential)
- Emits lifecycle events
- Handles steering/follow-up messages
**Key functions**:
- `agentLoop()` - Start new conversation
- `agentLoopContinue()` - Continue conversation
- `runAgentLoop()` - Internal loop execution
- `streamAssistantResponse()` - LLM API call
- `executeToolCalls()` - Tool execution
### Session
**What it is**: Conversation history management
**What it does**:
- Persists messages to storage
- Supports branching
- Implements compaction
- Manages conversation tree
**Key methods**:
- `appendMessage()` - Add message
- `appendCompaction()` - Compress history with summary
- `moveTo()` - Navigate branches
- `buildContext()` - Build context for LLM
- `getBranch()` - Get branch entries
- `getSessionStats()` - Get session statistics
- `appendThinkingLevelChange()` - Record thinking level change
- `appendModelChange()` - Record model change
- `appendActiveToolsChange()` - Record active tools change
### Tools
**What it is**: Functions agents can call
**What they do**:
- Execute external operations
- Return results to agent
- Support streaming updates
- Implement hooks
**Built-in tools**:
- `bash` - Execute shell commands
- `read` - Read files
- `write` - Write files
- `edit` - Edit files
## Event System
### Event Types
```
AgentEvent
├─ AgentStartEvent / AgentEndEvent
├─ TurnStartEvent / TurnEndEvent
├─ MessageStartEvent / MessageEndEvent
├─ MessageUpdateEvent
├─ ToolExecutionStartEvent / ToolExecutionEndEvent
└─ ToolExecutionUpdateEvent
```
### Event Flow
```
AgentStartEvent
├─ TurnStartEvent
│ ├─ MessageStartEvent (user)
│ ├─ MessageEndEvent (user)
│ ├─ MessageStartEvent (assistant)
│ ├─ MessageUpdateEvent (streaming)
│ ├─ MessageEndEvent (assistant)
│ ├─ ToolExecutionStartEvent
│ ├─ ToolExecutionEndEvent
│ └─ TurnEndEvent
└─ AgentEndEvent
```
## Complete Data Flow with Type Transformations
This documentation shows how data is transformed through the agent lifecycle.
### Message Type Hierarchy
```
Message (for LLM API)
├── UserMessage (role: "user")
│ └── content::Vector{MessageContent}
│ ├── TextContent (text::String)
│ └── ImageContent (data::String, mime_type::String)
├── AssistantMessage (role: "assistant")
│ ├── content::Vector{MessageContent}
│ │ ├── TextContent
│ │ └── ToolCall (id, name, arguments::Dict{String, Any})
│ ├── usage::Usage
│ ├── stop_reason::String
│ └── timestamp::Timestamp
└── ToolResultMessage (role: "toolResult")
├── tool_call_id::String
├── tool_name::String
├── content::Vector{MessageContent}
├── details::Any
├── usage::Union{Usage, Nothing}
├── is_error::Bool
└── timestamp::Timestamp
AgentMessage (internal, abstract type)
├── UserMessage (same as above)
├── AssistantMessage (same as above)
├── ToolResultMessage (same as above, plus: role, added_tool_names)
├── BashExecutionMessage (custom)
│ ├── role, command, output, exit_code
│ ├── cancelled, truncated, full_output_path, timestamp
│ └── exclude_from_context
├── CompactionSummaryMessage (custom)
│ ├── role, summary, tokens_before, timestamp
│ └── converted to UserMessage for LLM
├── BranchSummaryMessage (custom)
│ ├── role, summary, from_id, timestamp
│ └── converted to UserMessage for LLM
└── CustomMessage (custom, extends AgentMessage)
├── message::AgentMessage
└── custom_type::String
```
### Complete Conversation Flow
```
┌─────────────────────────────────────────────────────────────────────────────┐
│ Step 1: User Input (Vector{AgentMessage}) │
└─────────────────────────────────────────────────────────────────────────────┘
prompt(agent, "Hello!")
└─► normalizePromptInput()
Input: "Hello!"::String
Output: [UserMessage("user", [TextContent("Hello!")], timestamp)]
┌─────────────────────────────────────────────────────────────────────────────┐
│ Step 2: AgentLoop Processing │
└─────────────────────────────────────────────────────────────────────────────┘
runAgentLoop()
├─► transform_context() (optional hook)
│ Input: [UserMessage(...)]::Vector{AgentMessage}
│ Output: [UserMessage(...)]::Vector{AgentMessage}
├─► convert_to_llm()
│ Input: [UserMessage(...)]::Vector{AgentMessage}
│ Output: [UserMessage(...)]::Vector{Message}
├─► stream_fn() - LLM API call
│ Input: model, Context(...), config
│ Output: AssistantMessage with ToolCall[]
├─► executeToolCalls()
│ Input: AssistantMessage (with ToolCall[])
│ Output: ToolResultMessage[]
└─► Emit events and append to context.messages
┌─────────────────────────────────────────────────────────────────────────────┐
│ Step 3: Final Conversation State │
└─────────────────────────────────────────────────────────────────────────────┘
context.messages::Vector{AgentMessage}
├─ UserMessage("user", [TextContent("Hello!")], ...)
├─ AssistantMessage("assistant", [
│ TextContent("Hi there!"),
│ ToolCall("bash", {...})
│ ], ...)
└─ ToolResultMessage("toolResult", "bash", [TextContent("...")], ...)
┌─────────────────────────────────────────────────────────────────────────────┐
│ Step 4: AgentEndEvent (final output) │
└─────────────────────────────────────────────────────────────────────────────┘
AgentEndEvent(messages::Vector{AgentMessage})
└─ Contains full conversation history
User Input (String / AgentMessage / Vector{AgentMessage})
├─► normalizePromptInput()
│ Input: input::Union{String, AgentMessage, Vector{AgentMessage}}
│ Output: Vector{AgentMessage}
│ • String → UserMessage("user", [TextContent(input)], timestamp)
│ • AgentMessage → [input]
│ • Vector{AgentMessage} → input (pass-through)
├─► prompt(agent, messages)
│ └─► runPromptMessages()
AgentLoop Execution:
├─► transform_context() (optional hook)
│ Input: context.messages::Vector{AgentMessage}
│ Output: messages::Vector{AgentMessage} (transformed)
├─► convert_to_llm()
│ Input: messages::Vector{AgentMessage}
│ Output: llm_messages::Vector{Message}
│ AgentMessage → Message mapping:
│ • UserMessage → UserMessage (pass-through)
│ • AssistantMessage → AssistantMessage (pass-through)
│ • ToolResultMessage → ToolResultMessage (pass-through)
│ • BashExecutionMessage → UserMessage (text conversion)
│ • CompactionSummaryMessage → UserMessage (text wrapped)
│ • BranchSummaryMessage → UserMessage (text wrapped)
├─► LLM API Call (stream_fn)
│ Input: model, Context(system_prompt, llm_messages, tools), config
│ Output: Stream{AssistantMessageEvent}
├─► AssistantMessage (returned from LLM)
│ content::Vector{MessageContent}
│ └─ Contains: TextContent[] and/or ToolCall[]
├─► executeToolCalls() (if ToolCall[] in content)
│ │
│ ├─► prepareToolCall() for each ToolCall
│ │ Input: tool_call::ToolCall
│ │ Output: PreparedToolCall or ImmediateToolCallOutcome
│ │
│ ├─► executePreparedToolCall() (if prepared)
│ │ Input: PreparedToolCall
│ │ Output: ExecutedToolCallOutcome
│ │ tool.execute() returns AgentToolResultMutable
│ │
│ ├─► finalizeExecutedToolCall()
│ │ Input: ExecutedToolCallOutcome
│ │ Output: FinalizedToolCallOutcome
│ │
│ └─► createToolResultMessage()
│ Input: FinalizedToolCallOutcome
│ Output: ToolResultMessage
│ • role: "toolResult"
│ • tool_call_id, tool_name
│ • content::Vector{MessageContent}
│ • details, usage, added_tool_names
│ • is_error, timestamp
└─► Append to context.messages and new_messages
Vector{AgentMessage} (final conversation history)
Contains: [UserMessage, AssistantMessage, ToolResultMessage, ...]
```
### Tool Execution Flow
```
ToolCall (in AssistantMessage.content)
├─ before_tool_call hook (optional)
│ Input: BeforeToolCallContext
│ Output: BeforeToolCallResult (block, reason) or nothing
├─ prepareToolCall()
│ Input: tool_call::ToolCall
│ Output: Union{PreparedToolCall, ImmediateToolCallOutcome}
│ • Validates tool exists
│ • Runs before_tool_call hook
│ • Runs prepare_arguments hook (optional)
│ • Runs validateToolArguments (optional)
├─ executePreparedToolCall() (if prepared)
│ Input: PreparedToolCall
│ Output: ExecutedToolCallOutcome
│ tool.execute() returns AgentToolResultMutable
├─ finalizeExecutedToolCall()
│ Input: ExecutedToolCallOutcome
│ Output: FinalizedToolCallOutcome
│ Runs after_tool_call hook (optional)
└─ createToolResultMessage()
Input: FinalizedToolCallOutcome
Output: ToolResultMessage
• role: "toolResult"
• tool_call_id, tool_name
• content::Vector{MessageContent}
• details, usage, added_tool_names
• is_error, timestamp
```
## Best Practices
### 1. Use Hooks for Customization
```julia
# Before tool call
before_hook = (context, signal) -> begin
println("Executing: $(context.tool_call.name)")
return nothing
end
# After tool call
after_hook = (context, signal) -> begin
if context.is_error
println("Tool failed: $(context.tool_call.name)")
end
return nothing
end
```
### 2. Monitor Events
```julia
subscribe(agent) do event, signal
if event isa MessageEndEvent
println("Message: $(event.message)")
elseif event isa ToolExecutionEndEvent
println("Tool completed: $(event.tool_name)")
end
end
```
### 3. Use Steering for Redirection
```julia
# Agent is going wrong direction
steer(agent, UserMessage("Actually, let's do X instead"))
```
### 4. Use Follow-Up for Continuation
```julia
# Agent thinks it's done, but user wants more
followUp(agent, UserMessage("Wait, there's one more thing"))
```
## Common Patterns
### Pattern 1: Conversation with Memory
```julia
# Use Session to persist conversation
storage = JsonlSessionStorage(...)
session = Session(storage)
# Add messages to session
appendMessage(session, user_message)
appendMessage(session, assistant_message)
# Build context from session
context = buildContext(session)
```
### Pattern 2: Long Conversations
```julia
# Compact periodically to stay within context limits
if token_count > MAX_TOKENS * 0.8
compact_id = appendCompaction(
session,
summary,
first_kept_id,
token_count,
)
end
```
### Pattern 3: Branching Conversations
```julia
# User wants to explore alternative
session.moveTo(branch_point_id)
# Create new branch
moveTo(session, branch_point_id, summary=["summary" => "Exploring alternative approach"])
appendMessage(session, new_user_message)
```
### Pattern 4: Custom Tools
```julia
# Create custom tool
custom_tool = AgentTool(
"custom", # name
"Custom", # label
"Does custom thing", # description
parameters, # parameter schema
execute_function, # execute
nothing, # prepare_arguments (optional)
EXECUTION_PARALLEL, # execution_mode
)
# Add to agent
agent = Agent(Dict(:tools => [custom_tool]))
```
## Debugging
### Check Active Run
```julia
if !isnothing(agent.active_run)
println("Agent is busy")
else
println("Agent is idle")
end
```
### Clear Queues
```julia
clearAllQueues(agent)
```
### Reset State
```julia
reset!(agent)
```
## Performance Tips
1. **Use parallel execution** for independent tools
2. **Compact periodically** for long conversations
3. **Use thinking_level wisely** (higher = slower but better)
4. **Batch tool calls** when possible
5. **Cache LLM responses** when appropriate
## Troubleshooting
### Agent stuck in loop
```julia
# Check if agent is still processing
if hasQueuedMessages(agent)
# Clear queues
clearAllQueues(agent)
end
```
### Too many tokens
```julia
# Compact session
compact_id = appendCompaction(
session,
summary,
first_kept_id,
token_count,
)
```
### Tool execution failed
```julia
# Check tool result
if result.is_error
println("Tool failed: $(result.error)")
end
```
## Next Steps
1. Read **Architecture Overview** for deep understanding
2. Explore **Agent Component** for state management
3. Study **AgentLoop** for core logic
4. Learn **Types & Messages** for data structures
5. Master **Session Management** for persistence
6. Build **Tools** for custom functionality
## Resources
- Original TypeScript implementation: `@earendil-works/pi-agent-core`
- AgentCore.jl source code: `src/`
- Examples: `examples/`
## Community
For questions and discussions:
- GitHub Issues: `/issues`
- Documentation: `docs/`