hypoforge-python/config.toml at main · prudhvi1709/hypoforge-python · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
# HypoForge Configuration

[app]
title = "Hypothesis Forge"
version = "1.0.0"
port = 8000
host = "0.0.0.0"

[defaults]
api_base_url = "https://llmfoundry.straive.com/openai/v1"
model_name = "gpt-4.1-nano"
temperature = 0
max_age_hours = 24

[prompts]
hypothesis_default = "You are an expert data analyst. Generate hypotheses that would be valuable to test on this dataset. Each hypothesis should be clear, specific, and testable."

analysis_default = """You are an expert data scientist. Given a hypothesis and dataset description, write Python code to test this hypothesis using appropriate statistical methods.

CRITICAL: Your code must follow this exact pattern:

```python
import pandas as pd
import scipy.stats as stats
import numpy as np

def test_hypothesis(df):
    # Your analysis code here
    # Use appropriate statistical tests (t-test, chi-square, correlation, etc.)
    # Calculate p-value
    # Return (is_significant, p_value) where is_significant is boolean

    # Example:
    # statistic, p_value = stats.ttest_ind(group1, group2)
    # return p_value < 0.05, p_value

    pass
```

Use the most appropriate statistical test based on the hypothesis and data types. Always return exactly (boolean, float)."""

summary_system = """You are an expert data analyst.
Given a hypothesis and its outcome, provide a plain English summary of the findings as a crisp H5 heading (#####), followed by 1-2 concise supporting sentences.
Highlight in **bold** the keywords in the supporting statements.
Do not mention the p-value but _interpret_ it to support the conclusion quantitatively."""

synthesis_system = """Given the below hypotheses and results, summarize the key takeaways and actions in Markdown.
Begin with the hypotheses with lowest p-values AND highest business impact. Ignore results with errors.
Use action titles has H5 (#####). Just reading titles should tell the audience EXACTLY what to do.
Below each, add supporting bullet points that
  - PROVE the action title, mentioning which hypotheses led to this conclusion.
  - Do not mention the p-value but _interpret_ it to support the action
  - Highlight key phrases in **bold**.
Finally, after a break (---) add a 1-paragraph executive summary section (H5) summarizing these actions."""

[json_schema]
name = "HypothesesResponse"

[json_schema.schema]
type = "object"
required = ["hypotheses"]

[json_schema.schema.properties.hypotheses]
type = "array"

[json_schema.schema.properties.hypotheses.items]
type = "object"
required = ["hypothesis", "benefit"]

[json_schema.schema.properties.hypotheses.items.properties.hypothesis]
type = "string"

[json_schema.schema.properties.hypotheses.items.properties.benefit]
type = "string"