Compare commits
10 Commits
d46138290a
...
b91879a59e
| Author | SHA1 | Date |
|---|---|---|
|
|
b91879a59e | |
|
|
23e36bfbdf | |
|
|
129b9b07c7 | |
|
|
9c3def9884 | |
|
|
fad4ed1717 | |
|
|
f98e9f4806 | |
|
|
66d2fa09f9 | |
|
|
7dd0fc4a11 | |
|
|
fd5817fdcc | |
|
|
3f1f1d6573 |
134
.cursorrules
134
.cursorrules
|
|
@ -156,10 +156,119 @@ my_project/
|
|||
└── design.md
|
||||
```
|
||||
|
||||
- **`requirements.txt`**: Lists the Python dependencies for the project.
|
||||
```
|
||||
PyYAML
|
||||
pocketflow
|
||||
```
|
||||
|
||||
- **`docs/design.md`**: Contains project documentation for each step above. This should be *high-level* and *no-code*.
|
||||
~~~
|
||||
# Design Doc: Your Project Name
|
||||
|
||||
> Please DON'T remove notes for AI
|
||||
|
||||
## Requirements
|
||||
|
||||
> Notes for AI: Keep it simple and clear.
|
||||
> If the requirements are abstract, write concrete user stories
|
||||
|
||||
|
||||
## Flow Design
|
||||
|
||||
> Notes for AI:
|
||||
> 1. Consider the design patterns of agent, map-reduce, rag, and workflow. Apply them if they fit.
|
||||
> 2. Present a concise, high-level description of the workflow.
|
||||
|
||||
### Applicable Design Pattern:
|
||||
|
||||
1. Map the file summary into chunks, then reduce these chunks into a final summary.
|
||||
2. Agentic file finder
|
||||
- *Context*: The entire summary of the file
|
||||
- *Action*: Find the file
|
||||
|
||||
### Flow high-level Design:
|
||||
|
||||
1. **First Node**: This node is for ...
|
||||
2. **Second Node**: This node is for ...
|
||||
3. **Third Node**: This node is for ...
|
||||
|
||||
```mermaid
|
||||
flowchart TD
|
||||
firstNode[First Node] --> secondNode[Second Node]
|
||||
secondNode --> thirdNode[Third Node]
|
||||
```
|
||||
## Utility Functions
|
||||
|
||||
> Notes for AI:
|
||||
> 1. Understand the utility function definition thoroughly by reviewing the doc.
|
||||
> 2. Include only the necessary utility functions, based on nodes in the flow.
|
||||
|
||||
1. **Call LLM** (`utils/call_llm.py`)
|
||||
- *Input*: prompt (str)
|
||||
- *Output*: response (str)
|
||||
- Generally used by most nodes for LLM tasks
|
||||
|
||||
2. **Embedding** (`utils/get_embedding.py`)
|
||||
- *Input*: str
|
||||
- *Output*: a vector of 3072 floats
|
||||
- Used by the second node to embed text
|
||||
|
||||
## Node Design
|
||||
|
||||
### Shared Store
|
||||
|
||||
> Notes for AI: Try to minimize data redundancy
|
||||
|
||||
The shared store structure is organized as follows:
|
||||
|
||||
```python
|
||||
shared = {
|
||||
"key": "value"
|
||||
}
|
||||
```
|
||||
|
||||
### Node Steps
|
||||
|
||||
> Notes for AI: Carefully decide whether to use Batch/Async Node/Flow.
|
||||
|
||||
1. First Node
|
||||
- *Purpose*: Provide a short explanation of the node’s function
|
||||
- *Type*: Decide between Regular, Batch, or Async
|
||||
- *Steps*:
|
||||
- *prep*: Read "key" from the shared store
|
||||
- *exec*: Call the utility function
|
||||
- *post*: Write "key" to the shared store
|
||||
|
||||
2. Second Node
|
||||
...
|
||||
~~~
|
||||
|
||||
|
||||
- **`utils/`**: Contains all utility functions.
|
||||
- It's recommended to dedicate one Python file to each API call, for example `call_llm.py` or `search_web.py`.
|
||||
- Each file should also include a `main()` function to try that API call
|
||||
```python
|
||||
from google import genai
|
||||
import os
|
||||
|
||||
def call_llm(prompt: str) -> str:
|
||||
client = genai.Client(
|
||||
api_key=os.getenv("GEMINI_API_KEY", ""),
|
||||
)
|
||||
model = os.getenv("GEMINI_MODEL", "gemini-2.5-flash")
|
||||
response = client.models.generate_content(model=model, contents=[prompt])
|
||||
return response.text
|
||||
|
||||
if __name__ == "__main__":
|
||||
test_prompt = "Hello, how are you?"
|
||||
|
||||
# First call - should hit the API
|
||||
print("Making call...")
|
||||
response1 = call_llm(test_prompt, use_cache=False)
|
||||
print(f"Response: {response1}")
|
||||
```
|
||||
|
||||
- **`nodes.py`**: Contains all the node definitions.
|
||||
```python
|
||||
# nodes.py
|
||||
|
|
@ -1559,24 +1668,25 @@ Here, we provide some minimal example implementations:
|
|||
def call_llm(prompt):
|
||||
from anthropic import Anthropic
|
||||
client = Anthropic(api_key="YOUR_API_KEY_HERE")
|
||||
response = client.messages.create(
|
||||
model="claude-2",
|
||||
messages=[{"role": "user", "content": prompt}],
|
||||
max_tokens=100
|
||||
r = client.messages.create(
|
||||
model="claude-sonnet-4-0",
|
||||
messages=[
|
||||
{"role": "user", "content": prompt}
|
||||
]
|
||||
)
|
||||
return response.content
|
||||
return r.content[0].text
|
||||
```
|
||||
|
||||
3. Google (Generative AI Studio / PaLM API)
|
||||
```python
|
||||
def call_llm(prompt):
|
||||
import google.generativeai as genai
|
||||
genai.configure(api_key="YOUR_API_KEY_HERE")
|
||||
response = genai.generate_text(
|
||||
model="models/text-bison-001",
|
||||
prompt=prompt
|
||||
)
|
||||
return response.result
|
||||
from google import genai
|
||||
client = genai.Client(api_key='GEMINI_API_KEY')
|
||||
response = client.models.generate_content(
|
||||
model='gemini-2.5-pro',
|
||||
contents=prompt
|
||||
)
|
||||
return response.text
|
||||
```
|
||||
|
||||
4. Azure (Azure OpenAI)
|
||||
|
|
|
|||
|
|
@ -84,3 +84,4 @@ pyproject.toml
|
|||
usage.md
|
||||
cookbook/pocketflow-minimal-example/viz/flow_visualization.html
|
||||
cookbook/pocketflow-minimal-example/viz/flow_visualization.json
|
||||
.claude/
|
||||
|
|
|
|||
|
|
@ -0,0 +1,76 @@
|
|||
"""Registry describing files the natural-language assistant can modify."""
|
||||
from __future__ import annotations
|
||||
|
||||
from dataclasses import dataclass
|
||||
from pathlib import Path
|
||||
from typing import Dict, List
|
||||
|
||||
BASE_DIR = Path(__file__).resolve().parent
|
||||
|
||||
|
||||
@dataclass(frozen=True)
|
||||
class Resource:
|
||||
key: str
|
||||
path: Path
|
||||
description: str
|
||||
kind: str # e.g., "text" or "calendar"
|
||||
keywords: List[str]
|
||||
|
||||
|
||||
RESOURCE_REGISTRY: Dict[str, Resource] = {
|
||||
"shopping_list": Resource(
|
||||
key="shopping_list",
|
||||
path=BASE_DIR / "shopping_list.txt",
|
||||
description="Items to purchase on the next store run.",
|
||||
kind="text",
|
||||
keywords=["buy", "purchase", "shop", "grocery", "groceries"],
|
||||
),
|
||||
"home_todo": Resource(
|
||||
key="home_todo",
|
||||
path=BASE_DIR / "home_todo.txt",
|
||||
description="Household maintenance or chores.",
|
||||
kind="text",
|
||||
keywords=["home", "house", "chore", "laundry", "clean"],
|
||||
),
|
||||
"work_todo": Resource(
|
||||
key="work_todo",
|
||||
path=BASE_DIR / "work_todo.txt",
|
||||
description="Tasks related to your job or ongoing projects.",
|
||||
kind="text",
|
||||
keywords=["work", "office", "project", "client", "email","gavilan", "efw", ],
|
||||
),
|
||||
"school_courses": Resource(
|
||||
key="school_courses",
|
||||
path=BASE_DIR / "school_courses.txt",
|
||||
description="Assignments or study notes for school work.",
|
||||
kind="text",
|
||||
keywords=["school", "class", "course", "study", "assignment"],
|
||||
),
|
||||
"ideas": Resource(
|
||||
key="ideas",
|
||||
path=BASE_DIR / "ideas.txt",
|
||||
description="Random ideas, inspiration, or brainstorming notes.",
|
||||
kind="text",
|
||||
keywords=["idea", "brainstorm", "concept", "inspiration"],
|
||||
),
|
||||
"calendar": Resource(
|
||||
key="calendar",
|
||||
path=BASE_DIR / "calendar.ics",
|
||||
description="Time-based events saved in an iCalendar file.",
|
||||
kind="calendar",
|
||||
keywords=["meeting", "schedule", "calendar", "appointment", "event"],
|
||||
),
|
||||
}
|
||||
|
||||
|
||||
def resource_descriptions() -> str:
|
||||
"""Return a human-readable summary of supported resources."""
|
||||
lines = [
|
||||
"The assistant can update the following resources:",
|
||||
]
|
||||
for res in RESOURCE_REGISTRY.values():
|
||||
lines.append(f"- {res.key}: {res.description} (file: {res.path.name})")
|
||||
return "\n".join(lines)
|
||||
|
||||
|
||||
__all__ = ["RESOURCE_REGISTRY", "resource_descriptions", "Resource"]
|
||||
|
|
@ -0,0 +1,4 @@
|
|||
BEGIN:VCALENDAR
|
||||
VERSION:2.0
|
||||
PRODID:-//PocketFlow//Assistant//EN
|
||||
END:VCALENDAR
|
||||
|
|
@ -1,5 +1,6 @@
|
|||
pocketflow>=0.0.1
|
||||
aiohttp>=3.8.0 # For HTTP requests
|
||||
openai>=1.0.0 # For LLM calls
|
||||
duckduckgo-search>=7.5.2 # For web search
|
||||
requests>=2.25.1 # For HTTP requests
|
||||
duckduckgo-search>=7.5.2 # For web search
|
||||
aiohttp>=3.8.0 # For HTTP requests
|
||||
openai>=1.0.0 # For LLM calls
|
||||
requests>=2.25.1 # For HTTP requests
|
||||
PyYAML>=6.0.2 # For YAML parsing
|
||||
|
|
@ -0,0 +1,73 @@
|
|||
from async_flow import *
|
||||
from pocketflow import Flow, AsyncParallelBatchNode, Node
|
||||
|
||||
# Create node instances
|
||||
validate_payment = ValidatePayment()
|
||||
process_payment = ProcessPayment()
|
||||
payment_confirmation = PaymentConfirmation()
|
||||
|
||||
check_stock = CheckStock()
|
||||
reserve_items = ReserveItems()
|
||||
update_inventory = UpdateInventory()
|
||||
|
||||
create_label = CreateLabel()
|
||||
assign_carrier = AssignCarrier()
|
||||
schedule_pickup = SchedulePickup()
|
||||
|
||||
# Payment processing sub-flow
|
||||
validate_payment >> process_payment
|
||||
validate_payment - "out_of_stock" >> validate_payment # 循环重试
|
||||
process_payment - 'something fail' >> validate_payment
|
||||
process_payment - 'pass' >> payment_confirmation
|
||||
payment_flow = AsyncFlow(start=validate_payment)
|
||||
|
||||
# Inventory sub-flow
|
||||
check_stock >> reserve_items >> update_inventory
|
||||
inventory_flow = AsyncFlow(start=check_stock)
|
||||
|
||||
# Shipping sub-flow
|
||||
create_label >> assign_carrier >> schedule_pickup
|
||||
shipping_flow = AsyncFlow(start=create_label)
|
||||
|
||||
# Connect the flows into a main order pipeline
|
||||
payment_flow >> inventory_flow >> shipping_flow
|
||||
# payment_flow >> inventory_flow >> create_label
|
||||
# payment_flow >> inventory_flow >> assign_carrier
|
||||
|
||||
|
||||
# Create the master flow
|
||||
class OrderFlow(AsyncFlow):
|
||||
pass
|
||||
|
||||
order_pipeline = OrderFlow(start=payment_flow)
|
||||
|
||||
# Create shared data structure
|
||||
shared_data = {
|
||||
"order_id": "ORD-12345",
|
||||
"customer": "John Doe",
|
||||
"items": [
|
||||
{"id": "ITEM-001", "name": "Smartphone", "price": 999.99, "quantity": 1},
|
||||
{"id": "ITEM-002", "name": "Phone case", "price": 29.99, "quantity": 1},
|
||||
],
|
||||
"shipping_address": {
|
||||
"street": "123 Main St",
|
||||
"city": "Anytown",
|
||||
"state": "CA",
|
||||
"zip": "12345",
|
||||
},
|
||||
}
|
||||
|
||||
|
||||
# Run the entire pipeline asynchronously
|
||||
async def main():
|
||||
await order_pipeline.run_async(shared_data)
|
||||
|
||||
# Print final status
|
||||
print("\nOrder processing completed!")
|
||||
print(f"Payment: {shared_data.get('payment_confirmation')}")
|
||||
print(f"Inventory: {shared_data.get('inventory_update')}")
|
||||
print(f"Shipping: {shared_data.get('pickup_status')}")
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
asyncio.run(main())
|
||||
|
|
@ -77,6 +77,7 @@ def flow_to_json(start):
|
|||
node_types = {}
|
||||
flow_nodes = {} # Keep track of flow nodes
|
||||
ctr = 1
|
||||
visited = set()
|
||||
|
||||
def get_id(n):
|
||||
nonlocal ctr
|
||||
|
|
@ -99,6 +100,9 @@ def flow_to_json(start):
|
|||
action: Action label on the edge from parent to this node
|
||||
"""
|
||||
node_id = get_id(node)
|
||||
if (node_id, action) in visited:
|
||||
return
|
||||
visited.add((node_id, action))
|
||||
|
||||
# Add node if not already in nodes list and not a Flow
|
||||
if not any(n["id"] == node_id for n in nodes) and not isinstance(node, Flow):
|
||||
|
|
@ -552,8 +556,38 @@ def create_d3_visualization(
|
|||
|
||||
// Update positions on each tick
|
||||
simulation.on("tick", () => {
|
||||
// Update links with straight lines
|
||||
// Update links with curved paths for bidirectional connections
|
||||
link.attr("d", d => {
|
||||
// Handle self-referencing links with a water-drop shape
|
||||
if (d.source === d.target) {
|
||||
const nodeX = d.source.x;
|
||||
const nodeY = d.source.y;
|
||||
const offsetX = 40;
|
||||
const offsetY = 10;
|
||||
const controlOffset = 50;
|
||||
|
||||
// Create a water-drop shaped path
|
||||
return `M ${nodeX},${nodeY - 5}
|
||||
C ${nodeX + controlOffset},${nodeY - 30}
|
||||
${nodeX + offsetX},${nodeY + offsetY}
|
||||
${nodeX},${nodeY}`;
|
||||
}
|
||||
|
||||
// Check if there's a reverse connection
|
||||
const isReverse = data.links.some(l =>
|
||||
l.source === d.target && l.target === d.source
|
||||
);
|
||||
|
||||
// If it's part of a bidirectional connection, curve the path
|
||||
if (isReverse) {
|
||||
const dx = d.target.x - d.source.x;
|
||||
const dy = d.target.y - d.source.y;
|
||||
const dr = Math.sqrt(dx * dx + dy * dy) * 0.9;
|
||||
|
||||
return `M${d.source.x},${d.source.y}A${dr},${dr} 0 0,1 ${d.target.x},${d.target.y}`;
|
||||
}
|
||||
|
||||
// For unidirectional connections, use straight lines
|
||||
return `M${d.source.x},${d.source.y} L${d.target.x},${d.target.y}`;
|
||||
});
|
||||
|
||||
|
|
@ -567,10 +601,57 @@ def create_d3_visualization(
|
|||
.attr("x", d => d.x)
|
||||
.attr("y", d => d.y);
|
||||
|
||||
// Position link labels at midpoint
|
||||
linkLabel
|
||||
.attr("x", d => (d.source.x + d.target.x) / 2)
|
||||
.attr("y", d => (d.source.y + d.target.y) / 2);
|
||||
// Position link labels with offset for bidirectional connections
|
||||
linkLabel.attr("x", d => {
|
||||
// Handle self-referencing links
|
||||
if (d.source === d.target) {
|
||||
return d.source.x + 30;
|
||||
}
|
||||
|
||||
// Check if there's a reverse connection
|
||||
const reverseLink = data.links.find(l =>
|
||||
l.source === d.target && l.target === d.source
|
||||
);
|
||||
|
||||
// If it's part of a bidirectional connection, offset the label
|
||||
if (reverseLink) {
|
||||
const dx = d.target.x - d.source.x;
|
||||
const dy = d.target.y - d.source.y;
|
||||
// Calculate perpendicular offset
|
||||
const length = Math.sqrt(dx * dx + dy * dy);
|
||||
const offsetX = -dy / length * 10; // Perpendicular offset
|
||||
|
||||
return (d.source.x + d.target.x) / 2 + offsetX;
|
||||
}
|
||||
|
||||
// For unidirectional connections, use midpoint
|
||||
return (d.source.x + d.target.x) / 2;
|
||||
})
|
||||
.attr("y", d => {
|
||||
// Handle self-referencing links
|
||||
if (d.source === d.target) {
|
||||
return d.source.y;
|
||||
}
|
||||
|
||||
// Check if there's a reverse connection
|
||||
const reverseLink = data.links.find(l =>
|
||||
l.source === d.target && l.target === d.source
|
||||
);
|
||||
|
||||
// If it's part of a bidirectional connection, offset the label
|
||||
if (reverseLink) {
|
||||
const dx = d.target.x - d.source.x;
|
||||
const dy = d.target.y - d.source.y;
|
||||
// Calculate perpendicular offset
|
||||
const length = Math.sqrt(dx * dx + dy * dy);
|
||||
const offsetY = dx / length * 10; // Perpendicular offset
|
||||
|
||||
return (d.source.y + d.target.y) / 2 + offsetY;
|
||||
}
|
||||
|
||||
// For unidirectional connections, use midpoint
|
||||
return (d.source.y + d.target.y) / 2;
|
||||
});
|
||||
|
||||
// Update group containers
|
||||
groupContainers.each(function(d) {
|
||||
|
|
@ -893,7 +974,7 @@ if __name__ == "__main__":
|
|||
|
||||
parser = argparse.ArgumentParser(description="Visualize a PocketFlow flow")
|
||||
parser.add_argument(
|
||||
"--module", default="async_flow", help="Module containing the flow"
|
||||
"--module", default="async_loop_flow", help="Module containing the flow"
|
||||
)
|
||||
parser.add_argument(
|
||||
"--flow", default="order_pipeline", help="Flow variable name in the module"
|
||||
|
|
|
|||
109
docs/guide.md
109
docs/guide.md
|
|
@ -156,10 +156,119 @@ my_project/
|
|||
└── design.md
|
||||
```
|
||||
|
||||
- **`requirements.txt`**: Lists the Python dependencies for the project.
|
||||
```
|
||||
PyYAML
|
||||
pocketflow
|
||||
```
|
||||
|
||||
- **`docs/design.md`**: Contains project documentation for each step above. This should be *high-level* and *no-code*.
|
||||
~~~
|
||||
# Design Doc: Your Project Name
|
||||
|
||||
> Please DON'T remove notes for AI
|
||||
|
||||
## Requirements
|
||||
|
||||
> Notes for AI: Keep it simple and clear.
|
||||
> If the requirements are abstract, write concrete user stories
|
||||
|
||||
|
||||
## Flow Design
|
||||
|
||||
> Notes for AI:
|
||||
> 1. Consider the design patterns of agent, map-reduce, rag, and workflow. Apply them if they fit.
|
||||
> 2. Present a concise, high-level description of the workflow.
|
||||
|
||||
### Applicable Design Pattern:
|
||||
|
||||
1. Map the file summary into chunks, then reduce these chunks into a final summary.
|
||||
2. Agentic file finder
|
||||
- *Context*: The entire summary of the file
|
||||
- *Action*: Find the file
|
||||
|
||||
### Flow high-level Design:
|
||||
|
||||
1. **First Node**: This node is for ...
|
||||
2. **Second Node**: This node is for ...
|
||||
3. **Third Node**: This node is for ...
|
||||
|
||||
```mermaid
|
||||
flowchart TD
|
||||
firstNode[First Node] --> secondNode[Second Node]
|
||||
secondNode --> thirdNode[Third Node]
|
||||
```
|
||||
## Utility Functions
|
||||
|
||||
> Notes for AI:
|
||||
> 1. Understand the utility function definition thoroughly by reviewing the doc.
|
||||
> 2. Include only the necessary utility functions, based on nodes in the flow.
|
||||
|
||||
1. **Call LLM** (`utils/call_llm.py`)
|
||||
- *Input*: prompt (str)
|
||||
- *Output*: response (str)
|
||||
- Generally used by most nodes for LLM tasks
|
||||
|
||||
2. **Embedding** (`utils/get_embedding.py`)
|
||||
- *Input*: str
|
||||
- *Output*: a vector of 3072 floats
|
||||
- Used by the second node to embed text
|
||||
|
||||
## Node Design
|
||||
|
||||
### Shared Store
|
||||
|
||||
> Notes for AI: Try to minimize data redundancy
|
||||
|
||||
The shared store structure is organized as follows:
|
||||
|
||||
```python
|
||||
shared = {
|
||||
"key": "value"
|
||||
}
|
||||
```
|
||||
|
||||
### Node Steps
|
||||
|
||||
> Notes for AI: Carefully decide whether to use Batch/Async Node/Flow.
|
||||
|
||||
1. First Node
|
||||
- *Purpose*: Provide a short explanation of the node’s function
|
||||
- *Type*: Decide between Regular, Batch, or Async
|
||||
- *Steps*:
|
||||
- *prep*: Read "key" from the shared store
|
||||
- *exec*: Call the utility function
|
||||
- *post*: Write "key" to the shared store
|
||||
|
||||
2. Second Node
|
||||
...
|
||||
~~~
|
||||
|
||||
|
||||
- **`utils/`**: Contains all utility functions.
|
||||
- It's recommended to dedicate one Python file to each API call, for example `call_llm.py` or `search_web.py`.
|
||||
- Each file should also include a `main()` function to try that API call
|
||||
```python
|
||||
from google import genai
|
||||
import os
|
||||
|
||||
def call_llm(prompt: str) -> str:
|
||||
client = genai.Client(
|
||||
api_key=os.getenv("GEMINI_API_KEY", ""),
|
||||
)
|
||||
model = os.getenv("GEMINI_MODEL", "gemini-2.5-flash")
|
||||
response = client.models.generate_content(model=model, contents=[prompt])
|
||||
return response.text
|
||||
|
||||
if __name__ == "__main__":
|
||||
test_prompt = "Hello, how are you?"
|
||||
|
||||
# First call - should hit the API
|
||||
print("Making call...")
|
||||
response1 = call_llm(test_prompt, use_cache=False)
|
||||
print(f"Response: {response1}")
|
||||
```
|
||||
|
||||
- **`nodes.py`**: Contains all the node definitions.
|
||||
```python
|
||||
# nodes.py
|
||||
|
|
|
|||
|
|
@ -33,8 +33,7 @@ Here, we provide some minimal example implementations:
|
|||
from anthropic import Anthropic
|
||||
client = Anthropic(api_key="YOUR_API_KEY_HERE")
|
||||
r = client.messages.create(
|
||||
model="claude-3-7-sonnet-20250219",
|
||||
max_tokens=3000,
|
||||
model="claude-sonnet-4-0",
|
||||
messages=[
|
||||
{"role": "user", "content": prompt}
|
||||
]
|
||||
|
|
@ -48,7 +47,7 @@ Here, we provide some minimal example implementations:
|
|||
from google import genai
|
||||
client = genai.Client(api_key='GEMINI_API_KEY')
|
||||
response = client.models.generate_content(
|
||||
model='gemini-2.0-flash-001',
|
||||
model='gemini-2.5-pro',
|
||||
contents=prompt
|
||||
)
|
||||
return response.text
|
||||
|
|
|
|||
|
|
@ -62,10 +62,10 @@ class AsyncNode(Node):
|
|||
async def exec_fallback_async(self,prep_res,exc): raise exc
|
||||
async def post_async(self,shared,prep_res,exec_res): pass
|
||||
async def _exec(self,prep_res):
|
||||
for i in range(self.max_retries):
|
||||
for self.cur_retry in range(self.max_retries):
|
||||
try: return await self.exec_async(prep_res)
|
||||
except Exception as e:
|
||||
if i==self.max_retries-1: return await self.exec_fallback_async(prep_res,e)
|
||||
if self.cur_retry==self.max_retries-1: return await self.exec_fallback_async(prep_res,e)
|
||||
if self.wait>0: await asyncio.sleep(self.wait)
|
||||
async def run_async(self,shared):
|
||||
if self.successors: warnings.warn("Node won't run successors. Use AsyncFlow.")
|
||||
|
|
|
|||
2
setup.py
2
setup.py
|
|
@ -2,7 +2,7 @@ from setuptools import setup, find_packages
|
|||
|
||||
setup(
|
||||
name="pocketflow",
|
||||
version="0.0.2",
|
||||
version="0.0.3",
|
||||
packages=find_packages(),
|
||||
author="Zachary Huang",
|
||||
author_email="zh2408@columbia.edu",
|
||||
|
|
|
|||
|
|
@ -0,0 +1,248 @@
|
|||
"""Natural-language assistant that routes commands to specific personal files."""
|
||||
from __future__ import annotations
|
||||
|
||||
import re
|
||||
from datetime import datetime, timedelta, date, timezone
|
||||
from pathlib import Path
|
||||
from typing import Any, Dict, Optional
|
||||
|
||||
from pocketflow import Flow, Node
|
||||
|
||||
from assistant_resources import RESOURCE_REGISTRY, resource_descriptions
|
||||
from utils.call_llm import call_llm_json
|
||||
|
||||
COMMAND_PROMPT = f"""
|
||||
You convert a user's short reminder into a JSON command so an agent can update files.
|
||||
{resource_descriptions()}
|
||||
|
||||
Always respond with a JSON object using this schema:
|
||||
{{
|
||||
"action": "append",
|
||||
"target": "<one of: {', '.join(RESOURCE_REGISTRY.keys())}>",
|
||||
"entry": "<text to append or store>",
|
||||
"metadata": {{ # optional object for structured info (e.g., calendar dates)
|
||||
...
|
||||
}}
|
||||
}}
|
||||
If you cannot determine a valid target, respond with:
|
||||
{{"action": "unknown", "target": "", "entry": "", "metadata": {{}}}}
|
||||
Make entries concise; when the target is calendar include any useful scheduling details inside metadata.
|
||||
User reminder: {{command}}
|
||||
""".strip()
|
||||
|
||||
REMOVE_PREFIXES = [
|
||||
"remember to",
|
||||
"please",
|
||||
"don't forget to",
|
||||
"dont forget to",
|
||||
"remind me to",
|
||||
"i need to",
|
||||
"i should",
|
||||
]
|
||||
|
||||
|
||||
class InterpretCommandNode(Node):
|
||||
"""Use an LLM (with fallback heuristics) to map a reminder to an actionable command."""
|
||||
|
||||
def prep(self, shared: Dict[str, Any]) -> str:
|
||||
return shared["command"]
|
||||
|
||||
def exec(self, command: str) -> Dict[str, Any]:
|
||||
try:
|
||||
return call_llm_json(COMMAND_PROMPT.format(command=command))
|
||||
except Exception:
|
||||
return self._fallback(command)
|
||||
|
||||
def post(self, shared: Dict[str, Any], prep_result: str, exec_result: Dict[str, Any]) -> str:
|
||||
shared["parsed_action"] = exec_result
|
||||
return exec_result.get("action", "unknown")
|
||||
|
||||
@staticmethod
|
||||
def _fallback(command: str) -> Dict[str, Any]:
|
||||
lowered = command.lower()
|
||||
target_key: Optional[str] = None
|
||||
highest_score = 0
|
||||
for key, resource in RESOURCE_REGISTRY.items():
|
||||
score = sum(1 for kw in resource.keywords if kw in lowered)
|
||||
if score > highest_score:
|
||||
highest_score = score
|
||||
target_key = key
|
||||
if not target_key:
|
||||
return {"action": "unknown", "target": "", "entry": "", "metadata": {}}
|
||||
|
||||
entry = InterpretCommandNode._extract_entry(command, target_key)
|
||||
metadata: Dict[str, Any] = {}
|
||||
if target_key == "calendar":
|
||||
date_match = re.search(r"(\d{4}-\d{2}-\d{2})", command)
|
||||
if date_match:
|
||||
metadata["date"] = date_match.group(1)
|
||||
return {
|
||||
"action": "append",
|
||||
"target": target_key,
|
||||
"entry": entry.strip(),
|
||||
"metadata": metadata,
|
||||
}
|
||||
|
||||
@staticmethod
|
||||
def _extract_entry(command: str, target_key: str) -> str:
|
||||
cleaned = InterpretCommandNode._remove_prefixes(command).strip()
|
||||
lowered = cleaned.lower()
|
||||
if target_key == "shopping_list":
|
||||
match = re.search(r"buy\s+([a-zA-Z0-9\s]+)", lowered)
|
||||
if match:
|
||||
return match.group(1)
|
||||
if target_key in {"home_todo", "work_todo", "school_courses"}:
|
||||
for verb in ("finish", "do", "complete", "send", "write", "study"):
|
||||
if lowered.startswith(f"{verb} "):
|
||||
return cleaned
|
||||
if target_key == "ideas":
|
||||
return cleaned
|
||||
if target_key == "calendar":
|
||||
return cleaned
|
||||
return cleaned
|
||||
|
||||
@staticmethod
|
||||
def _remove_prefixes(command: str) -> str:
|
||||
lowered = command.lower()
|
||||
for prefix in REMOVE_PREFIXES:
|
||||
if lowered.startswith(prefix):
|
||||
return command[len(prefix) :].strip()
|
||||
return command
|
||||
|
||||
|
||||
class AppendEntryNode(Node):
|
||||
"""Append entries to the resource indicated by the interpreter."""
|
||||
|
||||
def prep(self, shared: Dict[str, Any]) -> Dict[str, Any]:
|
||||
return shared["parsed_action"]
|
||||
|
||||
def exec(self, action: Dict[str, Any]) -> str:
|
||||
target_key = action.get("target", "")
|
||||
entry = (action.get("entry") or "").strip()
|
||||
metadata = action.get("metadata") or {}
|
||||
|
||||
if target_key not in RESOURCE_REGISTRY:
|
||||
raise ValueError(f"Unsupported target '{target_key}'")
|
||||
if not entry:
|
||||
raise ValueError("Empty entry cannot be appended")
|
||||
|
||||
resource = RESOURCE_REGISTRY[target_key]
|
||||
if resource.kind == "text":
|
||||
self._append_to_text(resource.path, entry)
|
||||
elif resource.kind == "calendar":
|
||||
self._append_to_calendar(resource.path, entry, metadata)
|
||||
else:
|
||||
raise ValueError(f"Unknown resource kind '{resource.kind}'")
|
||||
return f"{target_key}:{entry}"
|
||||
|
||||
def post(self, shared: Dict[str, Any], prep_result: Dict[str, Any], exec_result: str) -> None:
|
||||
shared.setdefault("updates", []).append(exec_result)
|
||||
|
||||
@staticmethod
|
||||
def _append_to_text(path: Path, entry: str) -> None:
|
||||
path.parent.mkdir(parents=True, exist_ok=True)
|
||||
with path.open("a", encoding="utf-8") as fh:
|
||||
fh.write(f"{entry}\n")
|
||||
|
||||
@staticmethod
|
||||
def _append_to_calendar(path: Path, summary: str, metadata: Dict[str, Any]) -> None:
|
||||
path.parent.mkdir(parents=True, exist_ok=True)
|
||||
if not path.exists():
|
||||
path.write_text("BEGIN:VCALENDAR\nVERSION:2.0\nPRODID:-//PocketFlow//Assistant//EN\nEND:VCALENDAR\n", encoding="utf-8")
|
||||
|
||||
content = path.read_text(encoding="utf-8").splitlines()
|
||||
if not content or content[-1].strip() != "END:VCALENDAR":
|
||||
content.append("END:VCALENDAR")
|
||||
|
||||
event_lines = AppendEntryNode._build_event(summary, metadata)
|
||||
# insert before END:VCALENDAR
|
||||
end_index = len(content) - 1
|
||||
content = content[:end_index] + event_lines + content[end_index:]
|
||||
path.write_text("\n".join(content) + "\n", encoding="utf-8")
|
||||
|
||||
@staticmethod
|
||||
def _build_event(summary: str, metadata: Dict[str, Any]) -> list[str]:
|
||||
now = datetime.now(timezone.utc)
|
||||
dtstamp = now.strftime("%Y%m%dT%H%M%SZ")
|
||||
dtstart, dtend = AppendEntryNode._resolve_calendar_times(metadata, now)
|
||||
description = metadata.get("notes") or summary
|
||||
return [
|
||||
"BEGIN:VEVENT",
|
||||
f"DTSTAMP:{dtstamp}",
|
||||
f"SUMMARY:{summary}",
|
||||
f"DTSTART;VALUE=DATE:{dtstart}",
|
||||
f"DTEND;VALUE=DATE:{dtend}",
|
||||
f"DESCRIPTION:{description}",
|
||||
"END:VEVENT",
|
||||
]
|
||||
|
||||
@staticmethod
|
||||
def _resolve_calendar_times(metadata: Dict[str, Any], fallback: datetime) -> tuple[str, str]:
|
||||
start = AppendEntryNode._parse_date(metadata.get("date") or metadata.get("start"))
|
||||
if start is None:
|
||||
start = fallback.date()
|
||||
end = AppendEntryNode._parse_date(metadata.get("end"))
|
||||
if end is None:
|
||||
end_date = start + timedelta(days=1)
|
||||
else:
|
||||
end_date = end
|
||||
return start.strftime("%Y%m%d"), end_date.strftime("%Y%m%d")
|
||||
|
||||
@staticmethod
|
||||
def _parse_date(value: Any) -> Optional[date]:
|
||||
if not value:
|
||||
return None
|
||||
if isinstance(value, datetime):
|
||||
return value.date()
|
||||
try:
|
||||
# Try ISO date first
|
||||
dt = datetime.fromisoformat(str(value))
|
||||
return dt.date()
|
||||
except ValueError:
|
||||
patterns = ["%Y-%m-%d", "%m/%d/%Y", "%b %d %Y"]
|
||||
text = str(value)
|
||||
for pattern in patterns:
|
||||
try:
|
||||
return datetime.strptime(text, pattern).date()
|
||||
except ValueError:
|
||||
continue
|
||||
return None
|
||||
|
||||
|
||||
class UnknownCommandNode(Node):
|
||||
def prep(self, shared: Dict[str, Any]) -> Dict[str, Any]:
|
||||
return shared.get("parsed_action", {})
|
||||
|
||||
def exec(self, action: Dict[str, Any]) -> None:
|
||||
raise ValueError(f"Unsupported command: {action}")
|
||||
|
||||
|
||||
def build_flow() -> Flow:
|
||||
interpreter = InterpretCommandNode()
|
||||
append_entry = AppendEntryNode()
|
||||
unknown = UnknownCommandNode()
|
||||
|
||||
flow = Flow(start=interpreter)
|
||||
interpreter - "append" >> append_entry
|
||||
interpreter - "unknown" >> unknown
|
||||
return flow
|
||||
|
||||
|
||||
def handle_command(command: str) -> Dict[str, Any]:
|
||||
shared: Dict[str, Any] = {"command": command}
|
||||
flow = build_flow()
|
||||
try:
|
||||
flow.run(shared)
|
||||
except ValueError as exc:
|
||||
shared["error"] = str(exc)
|
||||
return shared
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
import argparse
|
||||
|
||||
parser = argparse.ArgumentParser(description="Process a reminder and update the relevant file")
|
||||
parser.add_argument("command", help="Reminder text, e.g. 'email client tomorrow about proposal'")
|
||||
args = parser.parse_args()
|
||||
result = handle_command(args.command)
|
||||
print(result)
|
||||
|
|
@ -0,0 +1,2 @@
|
|||
apples
|
||||
cheese
|
||||
|
|
@ -0,0 +1,54 @@
|
|||
"""Minimal OpenAI Chat Completions helper used by Pocket Flow demos."""
|
||||
from __future__ import annotations
|
||||
|
||||
import json
|
||||
from typing import Iterable, Optional
|
||||
|
||||
try:
|
||||
from openai import OpenAI
|
||||
except ImportError: # pragma: no cover - optional dependency
|
||||
OpenAI = None # type: ignore[misc,assignment]
|
||||
|
||||
try:
|
||||
from llm_secrets import OPENAI_API_KEY
|
||||
except ImportError as exc:
|
||||
raise ImportError("Create llm_secrets.py with OPENAI_API_KEY before calling call_llm") from exc
|
||||
|
||||
_client: Optional[OpenAI] = None
|
||||
|
||||
|
||||
def _get_client() -> OpenAI:
|
||||
global _client
|
||||
if OpenAI is None: # type: ignore[truthy-function]
|
||||
raise RuntimeError("Install the 'openai' package to use call_llm")
|
||||
if _client is None:
|
||||
if not OPENAI_API_KEY or OPENAI_API_KEY == "REPLACE_WITH_YOUR_KEY":
|
||||
raise ValueError("Set OPENAI_API_KEY in llm_secrets.py before calling call_llm")
|
||||
_client = OpenAI(api_key=OPENAI_API_KEY)
|
||||
return _client
|
||||
|
||||
|
||||
def call_llm(messages: Iterable[dict] | str, model: str = "gpt-4o-mini") -> str:
|
||||
"""Send a prompt or list of chat messages to OpenAI and return the text reply."""
|
||||
client = _get_client()
|
||||
chat_messages = (
|
||||
[{"role": "user", "content": messages}]
|
||||
if isinstance(messages, str)
|
||||
else list(messages)
|
||||
)
|
||||
response = client.chat.completions.create(model=model, messages=chat_messages)
|
||||
message = response.choices[0].message.content or ""
|
||||
return message.strip()
|
||||
|
||||
|
||||
def call_llm_json(messages: Iterable[dict] | str, model: str = "gpt-4o-mini") -> dict:
|
||||
"""Convenience wrapper that expects a JSON object in the response."""
|
||||
raw = call_llm(messages, model=model)
|
||||
start = raw.find("{")
|
||||
end = raw.rfind("}")
|
||||
if start == -1 or end == -1:
|
||||
raise ValueError(f"LLM response does not contain JSON: {raw}")
|
||||
return json.loads(raw[start : end + 1])
|
||||
|
||||
|
||||
__all__ = ["call_llm", "call_llm_json"]
|
||||
|
|
@ -0,0 +1 @@
|
|||
gavilan: call ted
|
||||
Loading…
Reference in New Issue