Spaces:
Running
Running
Update visuals.py
Browse files- visuals.py +43 -20
visuals.py
CHANGED
|
@@ -10,22 +10,18 @@ def create_progress_chart(log_data):
|
|
| 10 |
if not log_data or "trace" not in log_data:
|
| 11 |
return None
|
| 12 |
|
| 13 |
-
# Filter for "attempt" steps to get the scores
|
| 14 |
attempts = [step for step in log_data["trace"] if step["step_type"] == "attempt"]
|
| 15 |
if not attempts:
|
| 16 |
return None
|
| 17 |
|
| 18 |
-
# Define the standard criteria
|
| 19 |
categories = ["Novelty", "Usefulness_Feasibility", "Flexibility", "Elaboration", "Cultural_Appropriateness"]
|
| 20 |
|
| 21 |
fig = go.Figure()
|
| 22 |
|
| 23 |
for i, attempt in enumerate(attempts):
|
| 24 |
scores = attempt.get("scores", {})
|
| 25 |
-
# Handle cases where scores might be missing (default to 0)
|
| 26 |
values = [scores.get(cat, 0) for cat in categories]
|
| 27 |
|
| 28 |
-
# Close the loop for radar chart (repeat first value at the end)
|
| 29 |
values += [values[0]]
|
| 30 |
radar_categories = categories + [categories[0]]
|
| 31 |
|
|
@@ -44,7 +40,7 @@ def create_progress_chart(log_data):
|
|
| 44 |
polar=dict(
|
| 45 |
radialaxis=dict(
|
| 46 |
visible=True,
|
| 47 |
-
range=[0, 5]
|
| 48 |
)),
|
| 49 |
showlegend=True,
|
| 50 |
title="Evolution of Solution Quality"
|
|
@@ -58,24 +54,15 @@ def create_calibration_table(log_data):
|
|
| 58 |
if not log_data or "trace" not in log_data:
|
| 59 |
return None
|
| 60 |
|
| 61 |
-
# Find the calibration step in the log
|
| 62 |
calibration_step = next((step for step in log_data["trace"] if step["step_type"] == "calibration"), None)
|
| 63 |
-
|
| 64 |
-
# If no calibration happened (e.g. Single Agent mode), return None
|
| 65 |
if not calibration_step or "details" not in calibration_step:
|
| 66 |
return None
|
| 67 |
|
| 68 |
-
# details is a list of dicts: {'role': 'Plant', 'llm': 'Gemini', 'score': {...}}
|
| 69 |
details = calibration_step["details"]
|
| 70 |
-
|
| 71 |
data = []
|
| 72 |
for item in details:
|
| 73 |
role = item["role"]
|
| 74 |
model = item["llm"]
|
| 75 |
-
|
| 76 |
-
# The 'score' field contains the full evaluation object.
|
| 77 |
-
# We need to extract the specific numeric score relevant to that Role.
|
| 78 |
-
# Plant -> Novelty, Implementer -> Feasibility, Monitor -> Cultural
|
| 79 |
score_data = item.get("score", {})
|
| 80 |
|
| 81 |
score = 0
|
|
@@ -89,12 +76,48 @@ def create_calibration_table(log_data):
|
|
| 89 |
|
| 90 |
data.append({"Role": role, "Model": model, "Score": score})
|
| 91 |
|
| 92 |
-
if not data:
|
| 93 |
-
return None
|
| 94 |
|
| 95 |
df = pd.DataFrame(data)
|
| 96 |
-
|
| 97 |
-
# Pivot the table: Rows = Roles, Columns = Models, Values = Scores
|
| 98 |
-
# This creates a nice comparison matrix
|
| 99 |
pivot_df = df.pivot(index="Role", columns="Model", values="Score").reset_index()
|
| 100 |
-
return pivot_df
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 10 |
if not log_data or "trace" not in log_data:
|
| 11 |
return None
|
| 12 |
|
|
|
|
| 13 |
attempts = [step for step in log_data["trace"] if step["step_type"] == "attempt"]
|
| 14 |
if not attempts:
|
| 15 |
return None
|
| 16 |
|
|
|
|
| 17 |
categories = ["Novelty", "Usefulness_Feasibility", "Flexibility", "Elaboration", "Cultural_Appropriateness"]
|
| 18 |
|
| 19 |
fig = go.Figure()
|
| 20 |
|
| 21 |
for i, attempt in enumerate(attempts):
|
| 22 |
scores = attempt.get("scores", {})
|
|
|
|
| 23 |
values = [scores.get(cat, 0) for cat in categories]
|
| 24 |
|
|
|
|
| 25 |
values += [values[0]]
|
| 26 |
radar_categories = categories + [categories[0]]
|
| 27 |
|
|
|
|
| 40 |
polar=dict(
|
| 41 |
radialaxis=dict(
|
| 42 |
visible=True,
|
| 43 |
+
range=[0, 5]
|
| 44 |
)),
|
| 45 |
showlegend=True,
|
| 46 |
title="Evolution of Solution Quality"
|
|
|
|
| 54 |
if not log_data or "trace" not in log_data:
|
| 55 |
return None
|
| 56 |
|
|
|
|
| 57 |
calibration_step = next((step for step in log_data["trace"] if step["step_type"] == "calibration"), None)
|
|
|
|
|
|
|
| 58 |
if not calibration_step or "details" not in calibration_step:
|
| 59 |
return None
|
| 60 |
|
|
|
|
| 61 |
details = calibration_step["details"]
|
|
|
|
| 62 |
data = []
|
| 63 |
for item in details:
|
| 64 |
role = item["role"]
|
| 65 |
model = item["llm"]
|
|
|
|
|
|
|
|
|
|
|
|
|
| 66 |
score_data = item.get("score", {})
|
| 67 |
|
| 68 |
score = 0
|
|
|
|
| 76 |
|
| 77 |
data.append({"Role": role, "Model": model, "Score": score})
|
| 78 |
|
| 79 |
+
if not data: return None
|
|
|
|
| 80 |
|
| 81 |
df = pd.DataFrame(data)
|
|
|
|
|
|
|
|
|
|
| 82 |
pivot_df = df.pivot(index="Role", columns="Model", values="Score").reset_index()
|
| 83 |
+
return pivot_df
|
| 84 |
+
|
| 85 |
+
def create_cost_summary(log_data):
|
| 86 |
+
"""
|
| 87 |
+
Creates a Markdown summary of costs based on usage data found in the log.
|
| 88 |
+
"""
|
| 89 |
+
if not log_data or "financial_report" not in log_data:
|
| 90 |
+
return "**No Financial Data Available**"
|
| 91 |
+
|
| 92 |
+
fin = log_data["financial_report"]
|
| 93 |
+
total = fin.get("total_cost", 0)
|
| 94 |
+
calib = fin.get("calibration_cost", 0)
|
| 95 |
+
gen = fin.get("generation_cost", 0)
|
| 96 |
+
|
| 97 |
+
# Calculate Total Tokens
|
| 98 |
+
total_input = sum(u.get("input", 0) for u in fin.get("usage_breakdown", []))
|
| 99 |
+
total_output = sum(u.get("output", 0) for u in fin.get("usage_breakdown", []))
|
| 100 |
+
|
| 101 |
+
# Model Usage Breakdown
|
| 102 |
+
models_used = {}
|
| 103 |
+
for u in fin.get("usage_breakdown", []):
|
| 104 |
+
m = u.get("model", "Unknown")
|
| 105 |
+
models_used[m] = models_used.get(m, 0) + 1
|
| 106 |
+
|
| 107 |
+
model_str = ", ".join([f"{k} ({v} calls)" for k,v in models_used.items()])
|
| 108 |
+
|
| 109 |
+
md = f"""
|
| 110 |
+
### 💰 Financial Intelligence Report
|
| 111 |
+
|
| 112 |
+
| **Category** | **Cost (USD)** | **Details** |
|
| 113 |
+
| :--- | :--- | :--- |
|
| 114 |
+
| **Total Investment** | **${total:.6f}** | **Total execution cost** |
|
| 115 |
+
| Calibration Phase | ${calib:.6f} | Auditing models to pick the best team |
|
| 116 |
+
| Solution Phase | ${gen:.6f} | Drafting, refining, and judging |
|
| 117 |
+
|
| 118 |
+
---
|
| 119 |
+
**Operational Metrics:**
|
| 120 |
+
* **Total Tokens:** {total_input + total_output:,} ({total_input:,} in / {total_output:,} out)
|
| 121 |
+
* **Models Deployed:** {model_str}
|
| 122 |
+
"""
|
| 123 |
+
return md
|