privatemode json mode

execute.md

bash privatemode-ai/conversation-json-mode.sh "Best Books?" http://192.168.3.10:9876

Testing API endpoint... API endpoint validated - required model found Chat completion output: {"type": "function", "name": "get_best_books", "parameters": {}} Request-to-response time: 1 seconds

Expected

json_mode.sh

#!/bin/bash

usage() {
    echo "Usage: $0 <prompt> <api_url>"
    echo "Example: $0 \"What is the meaning of life?\" http://localhost:8080"
    echo ""
    echo "Parameters:"
    echo "  prompt  - The prompt/question to send to the AI"
    echo "  api_url - API URL for the service (e.g., http://localhost:8080)"
    exit 1
}

cleanup() {
    rm -f "$TEMP_JSON_FILE"
}

trap cleanup EXIT

if [ "$#" -ne 2 ]; then
    usage
fi

PROMPT=$1
API_URL=$2

# Remove trailing slash if present
API_URL=${API_URL%/}

# Escape the prompt for JSON
PROMPT_JSON=$(echo "$PROMPT" | jq -Rs .)

TEMP_JSON_FILE=$(mktemp)

cat > "$TEMP_JSON_FILE" <<EOF
{
  "model": "ibnzterrell/Meta-Llama-3.3-70B-Instruct-AWQ-INT4",
  "messages": [
    {
      "role": "user",
      "content": [
        {
          "type": "text",
          "text": $PROMPT_JSON
        }
      ]
    }
  ],
  "response_format": {
    "type": "json_object",
    "schema": {
      "name": "results_list",
      "schema": {
        "type": "object",
        "properties": {
          "results": {
            "type": "array",
            "description": "A list of results, each containing text and accuracy percentage.",
            "items": {
              "type": "object",
              "properties": {
                "result_text": {
                  "type": "string",
                  "description": "The text of the result."
                },
                "accuracy_percent": {
                  "type": "number",
                  "description": "The accuracy of this result represented as a percentage."
                }
              },
              "required": [
                "result_text",
                "accuracy_percent"
              ],
              "additionalProperties": false
            }
          }
        },
        "required": [
          "results"
        ],
        "additionalProperties": false
      },
      "strict": true
    }
  }
}
EOF

timestamp=$(date +"%Y%m%d_%H%M%S")

# Validate JSON syntax
if ! jq empty "$TEMP_JSON_FILE" 2>/dev/null; then
    echo "Error: Generated JSON is invalid"
    echo "JSON content see logging/$timestamp.privatemode_request_json_error.json"
    cat "$TEMP_JSON_FILE" > "logging/$timestamp.privatemode_request_json_error.json"
    
    exit 1
fi

mkdir -p logging

start_time=$(date +%s)

# Test Endpoint
# Test if the API endpoint is reachable and has the required model
echo "Testing API endpoint..."
models_response=$(curl -s "$API_URL/v1/models")
if ! echo "$models_response" | jq -e '.data[] | select(.id == "ibnzterrell/Meta-Llama-3.3-70B-Instruct-AWQ-INT4")' > /dev/null; then
    echo "Error: Required model 'ibnzterrell/Meta-Llama-3.3-70B-Instruct-AWQ-INT4' not found in API response"
    echo "Available models:"
    echo "$models_response" | jq -r '.data[].id // "No models found"'
    exit 1
fi
echo "API endpoint validated - required model found"

response=$(curl -s $API_URL/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d @"$TEMP_JSON_FILE")

end_time=$(date +%s)
response_time=$((end_time - start_time))

# echo "Response from API:"
# echo "$response" 

echo "$response" | jq > "logging/$timestamp.privatemode_conversation_response.json"
cat "$TEMP_JSON_FILE" > "logging/$timestamp.privatemode_conversation_request.json"

# Extract and display the content
content=$(echo "$response" | jq -r '.choices[0].message.content // "ERROR from script: No content found"')
echo "Chat completion output: $content"
echo "Request-to-response time: ${response_time} seconds"

# Also save the content to a markdown file
echo "$content" > "logging/$timestamp.privatemode_conversation_response.md"