Skip to content

Commit d1e3d00

Browse files
MervinPraisonMervinPraison
committed
Enhance Streamlit UIs and Add New Model Integrations
- Updated `mint.json` to include new Streamlit UIs for Deepseek, Gemini, and Ollama models, improving model accessibility. - Expanded documentation in `models.mdx` and `google.mdx` to provide detailed setup instructions for integrating OpenAI, Litellm, and Gemini models. - Introduced new example scripts for Gemini and Ollama Streamlit applications, showcasing interactive chat capabilities. - Added comprehensive features and prerequisites sections in the new UI documentation to guide users in setting up their environments. These changes enhance the usability and functionality of the PraisonAI framework, particularly in model integration and user interface development.
1 parent 969d2d0 commit d1e3d00

8 files changed

+754
-2
lines changed

‎docs/mint.json

+4-1
Original file line numberDiff line numberDiff line change
@@ -251,7 +251,10 @@
251251
"pages": [
252252
"ui/streamlit",
253253
"ui/streamlit/code-analysis-streamlit",
254-
"ui/streamlit/hackathon-judge-streamlit"
254+
"ui/streamlit/hackathon-judge-streamlit",
255+
"ui/streamlit/ollama-streamlit",
256+
"ui/streamlit/deepseek-streamlit",
257+
"ui/streamlit/gemini-streamlit"
255258
]
256259
},
257260
"ui/ui",

‎docs/models.mdx

+178-1
Original file line numberDiff line numberDiff line change
@@ -4,7 +4,184 @@ description: "Overview of supported language models in PraisonAI, including Open
44
icon: "brain"
55
---
66

7-
## Supported Models for Code
7+
# Code
8+
9+
## Set model by 3 ways
10+
11+
### 1. OpenAI Compatible Endpoints
12+
13+
<Note>By Default it uses OPENAI_BASE_URL https://api.openai.com/v1 </Note>
14+
Example Groq Implementation:
15+
16+
```bash
17+
export OPENAI_API_KEY=groq-api-key
18+
export OPENAI_BASE_URL=https://api.groq.com/openai/v1
19+
```
20+
21+
```python
22+
from praisonaiagents import Agent
23+
24+
agent = Agent(
25+
instructions="You are a helpful assistant",
26+
llm="llama-3.1-8b-instant",
27+
)
28+
29+
agent.start("Why sky is Blue?")
30+
```
31+
32+
### 2. Litellm Compatible model names (eg: gemini/gemini-1.5-flash-8b)
33+
34+
```bash
35+
pip install "praisonaiagents[llm]"
36+
```
37+
38+
```python
39+
from praisonaiagents import Agent
40+
41+
agent = Agent(
42+
instructions="You are a helpful assistant",
43+
llm="gemini/gemini-1.5-flash-8b",
44+
self_reflect=True,
45+
verbose=True
46+
)
47+
48+
agent.start("Why sky is Blue?")
49+
```
50+
51+
### 3. Litellm Compatible Configuration
52+
53+
```bash
54+
pip install "praisonaiagents[llm]"
55+
```
56+
57+
```python
58+
from praisonaiagents import Agent
59+
60+
llm_config = {
61+
"model": "gemini/gemini-1.5-flash-latest", # Model name without provider prefix
62+
63+
# Core settings
64+
"temperature": 0.7, # Controls randomness (like temperature)
65+
"timeout": 30, # Timeout in seconds
66+
"top_p": 0.9, # Nucleus sampling parameter
67+
"max_tokens": 1000, # Max tokens in response
68+
69+
# Advanced parameters
70+
"presence_penalty": 0.1, # Penalize repetition of topics (-2.0 to 2.0)
71+
"frequency_penalty": 0.1, # Penalize token repetition (-2.0 to 2.0)
72+
73+
# API settings (optional)
74+
"api_key": None, # Your API key (or use environment variable)
75+
"base_url": None, # Custom API endpoint if needed
76+
77+
# Response formatting
78+
"response_format": { # Force specific response format
79+
"type": "text" # Options: "text", "json_object"
80+
},
81+
82+
# Additional controls
83+
"seed": 42, # For reproducible responses
84+
"stop_phrases": ["##", "END"], # Custom stop sequences
85+
}
86+
87+
agent = Agent(
88+
instructions="You are a helpful Assistant."
89+
llm=llm_config
90+
)
91+
agent.start()
92+
```
93+
94+
## Advanced Configuration (Litellm Support)
95+
96+
<Note>This uses Litellm</Note>
97+
<Steps>
98+
<Step title="Install Package">
99+
Install required packages:
100+
```bash
101+
pip install "praisonaiagents[llm]"
102+
```
103+
</Step>
104+
105+
<Step title="Setup Environment">
106+
Configure environment:
107+
```bash
108+
export GOOGLE_API_KEY=your-api-key
109+
```
110+
111+
<Note>
112+
Get your API key from [Google AI Studio](https://makersuite.google.com/app/apikey)
113+
</Note>
114+
</Step>
115+
116+
<Step title="Create Agent">
117+
Create `app.py`:
118+
119+
<CodeGroup>
120+
121+
```python Basic
122+
# if json_object is supported by the model
123+
from praisonaiagents import Agent
124+
125+
agent = Agent(
126+
instructions="You are a helpful assistant",
127+
llm="gemini/gemini-1.5-flash-8b",
128+
self_reflect=True,
129+
verbose=True
130+
)
131+
132+
agent.start("Why sky is Blue?")
133+
```
134+
135+
```python Advanced
136+
# if json_object is not supported by the model
137+
from praisonaiagents import Agent
138+
139+
# Detailed LLM configuration
140+
llm_config = {
141+
"model": "gemini/gemini-1.5-flash-latest", # Model name without provider prefix
142+
143+
# Core settings
144+
"temperature": 0.7, # Controls randomness (like temperature)
145+
"timeout": 30, # Timeout in seconds
146+
"top_p": 0.9, # Nucleus sampling parameter
147+
"max_tokens": 1000, # Max tokens in response
148+
149+
# Advanced parameters
150+
"presence_penalty": 0.1, # Penalize repetition of topics (-2.0 to 2.0)
151+
"frequency_penalty": 0.1, # Penalize token repetition (-2.0 to 2.0)
152+
153+
# API settings (optional)
154+
"api_key": None, # Your API key (or use environment variable)
155+
"base_url": None, # Custom API endpoint if needed
156+
157+
# Response formatting
158+
"response_format": { # Force specific response format
159+
"type": "text" # Options: "text", "json_object"
160+
},
161+
162+
# Additional controls
163+
"seed": 42, # For reproducible responses
164+
"stop_phrases": ["##", "END"], # Custom stop sequences
165+
}
166+
167+
agent = Agent(
168+
instructions="You are a helpful Assistant specialized in scientific explanations. "
169+
"Provide clear, accurate, and engaging responses.",
170+
llm=llm_config, # Pass the detailed configuration
171+
verbose=True, # Enable detailed output
172+
markdown=True, # Format responses in markdown
173+
self_reflect=True, # Enable self-reflection
174+
max_reflect=3, # Maximum reflection iterations
175+
min_reflect=1 # Minimum reflection iterations
176+
)
177+
178+
# Test the agent
179+
response = agent.start("Why is the sky blue? Please explain in simple terms.")
180+
181+
```
182+
</CodeGroup>
183+
</Step>
184+
</Steps>
8185

9186
<AccordionGroup>
10187
<Accordion title="Ollama Integration" defaultOpen>

‎docs/models/google.mdx

+43
Original file line numberDiff line numberDiff line change
@@ -115,6 +115,49 @@ agents.start()
115115
This uses Litellm to connect to Google Gemini.
116116
</Note>
117117

118+
## Gemini 2.0 Flash Thinking
119+
120+
### LLM Configuration
121+
```python
122+
llm_config = {
123+
"model": "gemini/gemini-2.0-flash-thinking-exp-01-21",
124+
"response_format": {"type": "text"} # type is text, because json_object is not supported
125+
}
126+
```
127+
128+
<Steps>
129+
<Step title="Install Package">
130+
Install required packages:
131+
```bash
132+
pip install "praisonaiagents[llm]"
133+
```
134+
</Step>
135+
136+
<Step title="Set API Key">
137+
Set your Gemini API key:
138+
```bash
139+
export GEMINI_API_KEY=xxxxxxxxxxx
140+
```
141+
</Step>
142+
143+
<Step title="Code">
144+
145+
```python
146+
from praisonaiagents import Agent
147+
148+
llm_config = {"model": "gemini/gemini-2.0-flash-thinking-exp-01-21","response_format": {"type": "text"}}
149+
150+
agent = Agent(
151+
instructions="You are a helpful assistant",
152+
llm=llm_config
153+
)
154+
155+
result = agent.start("Why sky is Blue?")
156+
print(result)
157+
```
158+
</Step>
159+
</Steps>
160+
118161
## Alternative Setup
119162

120163
### Using Environment Variables

0 commit comments

Comments
 (0)