Change decision plugin output format from JSON to text

This commit is contained in:
Gigi
2025-04-04 18:15:53 +01:00
parent 09eec5607a
commit 11b46e485c

View File

@@ -3,7 +3,7 @@ description: Extract decisions made from the transcript
model: llama2 # Optional, falls back to OLLAMA_MODEL from .env
type: and # and, or
run: matching # always, matching
output_extension: .json
output_extension: .txt
prompt: |
You are a decision extraction specialist. Your job is to identify and extract decisions made during the conversation.
@@ -18,21 +18,17 @@ prompt: |
5. Any conditions or dependencies
Rules:
- Output should be in valid JSON format
- Each decision should be a separate object in an array
- Output should be in clear, readable text format
- Each decision should be separated by a blank line
- Include all relevant details for each decision
- Be specific and clear in the decision descriptions
- If certain information is not available, use null for that field
- If certain information is not available, indicate with "Not specified"
Format the output as:
{
"decisions": [
{
"decision": "string",
"made_by": "string or null",
"timestamp": "string or null",
"context": "string or null",
"conditions": "string or null"
}
]
}
Decision: [The decision made]
Made by: [Who made the decision or "Not specified"]
When: [When it was made or "Not specified"]
Context: [Context/reasoning or "Not specified"]
Conditions: [Conditions/dependencies or "Not specified"]
[Blank line between decisions]