Floke/Brancheneinstufung2
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix(gtm): Final robust prompt construction to eliminate SyntaxError and update migration guide

Browse Source
This commit is contained in:
Floke
2025-12-31 15:04:45 +00:00
parent 88e4a6d024
commit 9ad52893f9
2 changed files with 37 additions and 46 deletions
Download Patch File Download Diff File Expand all files Collapse all files

40
gtm_architect_orchestrator.py
View File

@@ -132,33 +132,33 @@ def analyze_product(data):
sys_instr = get_system_instruction(lang)
if lang == 'en':
extraction_prompt = f"""
extraction_prompt_template = """
PHASE 1-A: TECHNICAL EXTRACTION
Input Product Description: "{product_input[:25000]}..."
Input Product Description: "{product_description}"
Task:
1. Extract key technical features (specs, capabilities).
1) Extract key technical features (specs, capabilities).
2) Derive "Hard Constraints". IMPORTANT: Check Vmax (<20km/h = Private Grounds) and Cleaning Type (Vacuum != Heavy Debris/Snow).
3. Create a short raw analysis summary.
3) Create a short raw analysis summary.
Output JSON format ONLY.
"""
extraction_prompt = extraction_prompt_template.format(product_description=product_input[:25000])
else:
extraction_prompt = f"""
extraction_prompt_template = """
PHASE 1-A: TECHNICAL EXTRACTION
Input Product Description: "{product_input[:25000]}..."
Input Product Description: "{product_description}"
Aufgabe:
1. Extrahiere technische Hauptmerkmale (Specs, Fähigkeiten).
1) Extrahiere technische Hauptmerkmale (Specs, Fähigkeiten).
2) Leite "Harte Constraints" ab. WICHTIG: Prüfe Vmax (<20km/h = Privatgelände) und Reinigungstyp (Vakuum != Grobschmutz/Schnee).
3. Erstelle eine kurze Rohanalyse-Zusammenfassung.
3) Erstelle eine kurze Rohanalyse-Zusammenfassung.
Output JSON format ONLY.
"""
extraction_prompt = extraction_prompt_template.format(product_description=product_input[:25000])
# Fix: Prepend system instruction manually
full_extraction_prompt = sys_instr + "\n\n" + extraction_prompt
# Using response_format_json=True since helpers.py supports it (for logging/prompt hint)
extraction_response = call_openai_chat(full_extraction_prompt, response_format_json=True)
extraction_data = json.loads(extraction_response)
@@ -166,7 +166,7 @@ Output JSON format ONLY.
constraints_json = json.dumps(extraction_data.get('constraints'))
if lang == 'en':
conflict_prompt = f"""
conflict_prompt_template = """
PHASE 1-B: PORTFOLIO CONFLICT CHECK
New Product Features: {features_json}
@@ -181,8 +181,9 @@ Check if the new product overlaps significantly with existing ones (is it just a
Output JSON format ONLY.
"""
conflict_prompt = conflict_prompt_template.format(features_json=features_json, constraints_json=constraints_json)
else:
conflict_prompt = f"""
conflict_prompt_template = """
PHASE 1-B: PORTFOLIO CONFLICT CHECK
Neue Produkt-Features: {features_json}
@@ -197,8 +198,8 @@ Prüfe, ob das neue Produkt signifikant mit bestehenden Produkten überlappt (Is
Output JSON format ONLY.
"""
conflict_prompt = conflict_prompt_template.format(features_json=features_json, constraints_json=constraints_json)
# Fix: Prepend system instruction manually
full_conflict_prompt = sys_instr + "\n\n" + conflict_prompt
conflict_response = call_openai_chat(full_conflict_prompt, response_format_json=True)
conflict_data = json.loads(conflict_response)
@@ -215,7 +216,7 @@ def discover_icps(data):
constraints_json = json.dumps(phase1_result.get('constraints'))
if lang == 'en':
prompt = f"""
prompt_template = """
PHASE 2: ICP DISCOVERY & DATA PROXIES
Based on the product features: {features_json}
And constraints: {constraints_json}
@@ -226,8 +227,9 @@ Output JSON format ONLY:
"dataProxies": [ {{ "target": "Criteria", "method": "How" }} ]
}}
"""
prompt = prompt_template.format(features_json=features_json, constraints_json=constraints_json)
else:
prompt = f"""
prompt_template = """
PHASE 2: ICP DISCOVERY & DATA PROXIES
Basierend auf Features: {features_json}
Und Constraints: {constraints_json}
@@ -238,8 +240,8 @@ Output JSON format ONLY:
"dataProxies": [ {{ "target": "Kriterium", "method": "Methode" }} ]
}}
"""
prompt = prompt_template.format(features_json=features_json, constraints_json=constraints_json)
# Fix: Prepend system instruction manually
full_prompt = sys_instr + "\n\n" + prompt
response = call_openai_chat(full_prompt, response_format_json=True)
print(response)
@@ -250,7 +252,6 @@ def hunt_whales(data):
sys_instr = get_system_instruction(lang)
icps_json = json.dumps(phase2_result.get('icps'))
prompt = f"PHASE 3: WHALE HUNTING for {icps_json}. Identify 3-5 concrete DACH companies per industry and buying center roles. Output JSON ONLY."
# Fix: Prepend system instruction manually
full_prompt = sys_instr + "\n\n" + prompt
response = call_openai_chat(full_prompt, response_format_json=True)
print(response)
@@ -262,7 +263,6 @@ def develop_strategy(data):
sys_instr = get_system_instruction(lang)
phase3_json = json.dumps(phase3_result)
prompt = f"PHASE 4: STRATEGY Matrix for {phase3_json}. Apply Hybrid logic. Output JSON ONLY."
# Fix: Prepend system instruction manually
full_prompt = sys_instr + "\n\n" + prompt
response = call_openai_chat(full_prompt, response_format_json=True)
print(response)
@@ -272,7 +272,6 @@ def generate_assets(data):
sys_instr = get_system_instruction(lang)
data_json = json.dumps(data)
prompt = f"PHASE 5: GTM STRATEGY REPORT Markdown. Use facts, TCO, ROI. Data: {data_json}"
# Fix: Prepend system instruction manually
full_prompt = sys_instr + "\n\n" + prompt
response = call_openai_chat(full_prompt)
print(json.dumps(response))
@@ -282,7 +281,6 @@ def generate_sales_enablement(data):
sys_instr = get_system_instruction(lang)
data_json = json.dumps(data)
prompt = f"PHASE 6: Battlecards and MJ Prompts. Data: {data_json}. Output JSON ONLY."
# Fix: Prepend system instruction manually
full_prompt = sys_instr + "\n\n" + prompt
response = call_openai_chat(full_prompt, response_format_json=True)
print(response)
@@ -321,4 +319,4 @@ def main():
print(json.dumps({"error": f"Unknown mode: {args.mode}"}))
if __name__ == "__main__":
main()
main()