phowell
/
pocketflow
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

update structrued output demo

This commit is contained in:
zachary62 2025-04-18 19:14:59 -04:00
parent 9e026c7f14
commit e938bf3e42
2 changed files with 170 additions and 81 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

57
cookbook/pocketflow-structured-output/README.md
View File

@ -9,17 +9,27 @@ A minimal demo application showing how to use PocketFlow to extract structured d
## Run It ## Run It
1. Make sure your OpenAI API key is set: 1. Install the packages you need with this simple command:
```bash
pip install -r requirements.txt
```
2. Make sure your OpenAI API key is set:
```bash ```bash
export OPENAI_API_KEY="your-api-key-here" export OPENAI_API_KEY="your-api-key-here"
``` ```
Alternatively, you can edit the `utils.py` file to include your API key directly. Alternatively, you can edit the [`utils.py`](./utils.py) file to include your API key directly.
2. Edit data.txt with the resume you want to parse (a sample resume is already included) Let's do a quick check to make sure your API key is working properly:
3. Install requirements and run the application:
```bash ```bash
pip install -r requirements.txt python utils.py
```
3. Edit [data.txt](./data.txt) with the resume you want to parse (a sample resume is already included)
4. Run the application:
```bash
python main.py python main.py
``` ```
@ -45,24 +55,29 @@ The Resume Parser application uses a single node that:
## Example Output ## Example Output
``` ```
=== STRUCTURED RESUME DATA === === Resume Parser - Structured Output with Indexes & Comments ===
name: John Smith
=== STRUCTURED RESUME DATA (Comments & Skill Index List) ===
name: JOHN SMTIH
email: johnsmtih1983@gnail.com email: johnsmtih1983@gnail.com
experience: experience:
- title: Sales Manager - {title: SALES MANAGER, company: ABC Corportaion}
company: ABC Corporation - {title: ASST. MANAGER, company: XYZ Industries}
- title: Assistant Manager - {title: CUSTOMER SERVICE REPRESENTATIVE, company: Fast Solutions Inc}
company: XYZ Industries skill_indexes: [0, 1, 2, 3, 4]
- title: Customer Service Representative
company: Fast Solutions Inc
skills:
- Microsoft Office: Excel, Word, PowerPoint (Advanced)
- Customer relationship management (CRM) software
- Team leadership & management
- Project management
- Public speaking
- Time management
============================
============================================================
✅ Extracted resume information.
--- Found Target Skills (from Indexes) ---
- Team leadership & management (Index: 0)
- CRM software (Index: 1)
- Project management (Index: 2)
- Public speaking (Index: 3)
- Microsoft Office (Index: 4)
----------------------------------------
``` ```

194
cookbook/pocketflow-structured-output/main.py
View File

@ -1,74 +1,148 @@
from pocketflow import Node, Flow
from utils import call_llm
import yaml import yaml
import os # Needed for the utils import below
from pocketflow import Node, Flow
from utils import call_llm # Assumes utils.py with call_llm exists
class ResumeParserNode(Node): class ResumeParserNode(Node):
def prep(self, shared): def prep(self, shared):
"""Return resume text from shared state""" """Return resume text and target skills from shared state."""
return shared["resume_text"] return {
"resume_text": shared["resume_text"],
def exec(self, resume_text): "target_skills": shared.get("target_skills", [])
"""Extract structured data from resume using prompt engineering""" }
def exec(self, prep_res):
"""Extract structured data from resume using prompt engineering.
Requests YAML output with comments and skill indexes as a list.
"""
resume_text = prep_res["resume_text"]
target_skills = prep_res["target_skills"]
# Format skills with indexes for the prompt
skill_list_for_prompt = "\n".join([f"{i}: {skill}" for i, skill in enumerate(target_skills)])
# Simplified Prompt focusing on key instructions and format
prompt = f""" prompt = f"""
Please extract the following information from this resume and format it as YAML: Analyze the resume below. Output ONLY the requested information in YAML format.
- name
- email
- experience (list of positions with title and company)
- skills (list of skills)
**Resume:**
```
{resume_text} {resume_text}
```
Now, output: **Target Skills (use these indexes):**
```
{skill_list_for_prompt}
```
**YAML Output Requirements:**
- Extract `name` (string).
- Extract `email` (string).
- Extract `experience` (list of objects with `title` and `company`).
- Extract `skill_indexes` (list of integers found from the Target Skills list).
- **Add a YAML comment (`#`) explaining the source BEFORE `name`, `email`, `experience`, each item in `experience`, and `skill_indexes`.**
**Example Format:**
```yaml ```yaml
name: John Doe # Found name at top
email: john@example.com name: Jane Doe
# Found email in contact info
email: jane@example.com
# Experience section analysis
experience: experience:
- title: Software Engineer # First job listed
company: Tech Company - title: Manager
- title: Developer company: Corp A
company: Another Company # Second job listed
skills: - title: Assistant
- Python company: Corp B
- JavaScript # Skills identified from the target list based on resume content
- HTML/CSS skill_indexes:
```""" # Found 0 at top
- 0
# Found 2 in experience
- 2
```
Generate the YAML output now:
"""
response = call_llm(prompt) response = call_llm(prompt)
# Extract YAML content from markdown code blocks # --- Minimal YAML Extraction ---
# Assumes LLM correctly uses ```yaml blocks
yaml_str = response.split("```yaml")[1].split("```")[0].strip() yaml_str = response.split("```yaml")[1].split("```")[0].strip()
structured_result = yaml.safe_load(yaml_str) structured_result = yaml.safe_load(yaml_str)
# --- End Minimal Extraction ---
# Validate structure
assert "name" in structured_result
assert "experience" in structured_result
assert isinstance(structured_result["experience"], list)
assert "skills" in structured_result
assert isinstance(structured_result["skills"], list)
return structured_result
def post(self, shared, prep_res, exec_res):
"""Store and display structured resume data in YAML"""
shared["structured_data"] = exec_res
# Print structured data in YAML format
print("\n=== STRUCTURED RESUME DATA ===\n")
print(yaml.dump(exec_res, sort_keys=False))
print("\n============================\n")
print("✅ Extracted basic resume information")
# Create and run the flow
if __name__ == "__main__":
print("=== Simple Resume Parser - YAML Output ===\n")
# Read resume text from file
shared = {}
with open('data.txt', 'r') as file:
resume_text = file.read()
shared["resume_text"] = resume_text
# --- Basic Validation ---
flow = Flow(start=ResumeParserNode()) assert structured_result is not None, "Validation Failed: Parsed YAML is None"
flow.run(shared) assert "name" in structured_result, "Validation Failed: Missing 'name'"
assert "email" in structured_result, "Validation Failed: Missing 'email'"
assert "experience" in structured_result, "Validation Failed: Missing 'experience'"
assert isinstance(structured_result.get("experience"), list), "Validation Failed: 'experience' is not a list"
assert "skill_indexes" in structured_result, "Validation Failed: Missing 'skill_indexes'"
skill_indexes_val = structured_result.get("skill_indexes")
assert skill_indexes_val is None or isinstance(skill_indexes_val, list), "Validation Failed: 'skill_indexes' is not a list or None"
if isinstance(skill_indexes_val, list):
for index in skill_indexes_val:
assert isinstance(index, int), f"Validation Failed: Skill index '{index}' is not an integer"
# --- End Basic Validation ---
return structured_result
def post(self, shared, prep_res, exec_res):
"""Store structured data and print it."""
shared["structured_data"] = exec_res
print("\n=== STRUCTURED RESUME DATA (Comments & Skill Index List) ===\n")
# Dump YAML ensuring block style for readability
print(yaml.dump(exec_res, sort_keys=False, allow_unicode=True, default_flow_style=None))
print("\n============================================================\n")
print("✅ Extracted resume information.")
# === Main Execution Logic ===
if __name__ == "__main__":
print("=== Resume Parser - Structured Output with Indexes & Comments ===\n")
# --- Configuration ---
target_skills_to_find = [
"Team leadership & management", # 0
"CRM software", # 1
"Project management", # 2
"Public speaking", # 3
"Microsoft Office", # 4
"Python", # 5
"Data Analysis" # 6
]
resume_file = 'data.txt' # Assumes data.txt contains the resume
# --- Prepare Shared State ---
shared = {}
try:
with open(resume_file, 'r') as file:
shared["resume_text"] = file.read()
except FileNotFoundError:
print(f"Error: Resume file '{resume_file}' not found.")
exit(1) # Exit if resume file is missing
shared["target_skills"] = target_skills_to_find
# --- Define and Run Flow ---
parser_node = ResumeParserNode()
flow = Flow(start=parser_node)
flow.run(shared) # Execute the parsing node
# --- Display Found Skills ---
if "structured_data" in shared and "skill_indexes" in shared["structured_data"]:
print("\n--- Found Target Skills (from Indexes) ---")
found_indexes = shared["structured_data"]["skill_indexes"]
if found_indexes: # Check if the list is not empty or None
for index in found_indexes:
if 0 <= index < len(target_skills_to_find):
print(f"- {target_skills_to_find[index]} (Index: {index})")
else:
print(f"- Warning: Found invalid skill index {index}")
else:
print("No target skills identified from the list.")
print("----------------------------------------\n")