197 lines
8.5 KiB
Markdown
197 lines
8.5 KiB
Markdown
|
---
|
|||
|
|
license: apache-2.0
|
|||
|
|
language:
|
|||
|
|
- en
|
|||
|
|
pipeline_tag: text-generation
|
|||
|
|
library_name: transformers
|
|||
|
|
tags:
|
|||
|
|
- trl
|
|||
|
|
- text-generation-inference
|
|||
|
|
- agent
|
|||
|
|
- tool
|
|||
|
|
- web-agent
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
|
|||
|
|
|
|||
|
|
# **Dorado-WebSurf_Tool-ext**
|
|||
|
|
|
|||
|
|
> **Dorado-WebSurf_Tool-ext** is a **function-calling and agentic reasoning model** fine-tuned from **Qwen3-4B**, designed for **web search orchestration**, **tool-augmented reasoning**, and dynamic **problem-solving**.
|
|||
|
|
> It excels at **agentic decision-making**, **tool selection**, and structured execution flow, making it ideal for **retrieval-augmented generation (RAG)**, **function calling**, and **tool-based query resolution**.
|
|||
|
|
|
|||
|
|
> [!note]
|
|||
|
|
> GGUF: [https://huggingface.co/prithivMLmods/Dorado-WebSurf_Tool-ext-GGUF](https://huggingface.co/prithivMLmods/Dorado-WebSurf_Tool-ext-GGUF)
|
|||
|
|
|
|||
|
|
|
|||
|
|
## **Key Features**
|
|||
|
|
|
|||
|
|
1. **Agentic Reasoning & Tool-Oriented Execution**
|
|||
|
|
Built for orchestrating **function calls**, selecting and sequencing tools, and solving queries through structured multi-step reasoning.
|
|||
|
|
|
|||
|
|
2. **Web Search Query Orchestration**
|
|||
|
|
Integrates web search planning, retrieval grounding, and fact-checking, enabling intelligent **query resolution** from live data sources.
|
|||
|
|
|
|||
|
|
3. **Dynamic Tool Selection & Execution Chains**
|
|||
|
|
Chooses from an **array of available tools** — including web search, APIs, mathematical solvers, and structured data processors — to solve complex tasks.
|
|||
|
|
|
|||
|
|
4. **Hybrid Symbolic-Probabilistic Logic**
|
|||
|
|
Combines structured reasoning with probabilistic inference, ensuring accurate outcomes even in uncertainty-driven or multi-source contexts.
|
|||
|
|
|
|||
|
|
5. **Structured Output Generation**
|
|||
|
|
Generates responses in **JSON**, **YAML**, **Markdown**, or **tool call schema formats**, ideal for automation pipelines and agent frameworks.
|
|||
|
|
|
|||
|
|
6. **Optimized Lightweight Footprint**
|
|||
|
|
Maintains strong reasoning and tool orchestration capabilities in a **4B parameter model**, deployable on **mid-range GPUs**, **edge devices**, and **offline clusters**.
|
|||
|
|
|
|||
|
|
|
|||
|
|
# **Quick Start with Brave Search MCP**
|
|||
|
|
|
|||
|
|
mcp.json
|
|||
|
|
|
|||
|
|
```json
|
|||
|
|
{
|
|||
|
|
"mcpServers": {
|
|||
|
|
"brave-search": {
|
|||
|
|
"command": "npx",
|
|||
|
|
"args": [
|
|||
|
|
"-y",
|
|||
|
|
"@modelcontextprotocol/server-brave-search"
|
|||
|
|
],
|
|||
|
|
"env": {
|
|||
|
|
"BRAVE_API_KEY": "---Enter your Brave API key here.---",
|
|||
|
|
"BRAVE_API_URL": "https://api.search.brave.com/res/v1/web/search"
|
|||
|
|
}
|
|||
|
|
}
|
|||
|
|
}
|
|||
|
|
}
|
|||
|
|
```
|
|||
|
|
## Example : 01
|
|||
|
|
|
|||
|
|
Input Prompt:
|
|||
|
|
|
|||
|
|
```
|
|||
|
|
Asia Cup 2025 Final: India vs Pakistan – Match Results
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
```py
|
|||
|
|
|
|||
|
|
The search results indicate that there was an Asia Cup 2025 Final match between India and Pakistan, with **India beating Pakistan**. However, note that the Asia Cup is typically held every four years (e.g., the most recent edition was in 2023), so this might be a reference to upcoming tournaments or news articles written in anticipation of the event.
|
|||
|
|
|
|||
|
|
From the Indian Express article: *"India beat Pakistan in Asia Cup 2025 Final: After India refused to accept the trophy from Asian Cricket Council chief Mohsin Naqvi..."* suggests this match took place in reality (not just hypothetical).
|
|||
|
|
|
|||
|
|
I'd confirm via a quick check if this is real or for 2025 planning. Let's do an extra search to verify the article date.
|
|||
|
|
|
|||
|
|
Wait, but I need to return the final answer step here.
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
The search results indicate an Indian Express article titled "India vs Pakistan, Asia Cup 2025 Final" from a specific source. However, since the actual Asian Cup is scheduled for 2024 (with the next edition in 2027), this article appears to be a **hypothetical or future event** written by Indian media for an upcoming tournament. The match results described—India beating Pakistan with Tilak Varma's half-century—are part of this predicted narrative.
|
|||
|
|
|
|||
|
|
So, based on this fictional news item from The Indian Express (likely published in anticipation of the 2025 final), **India won against Pakistan in the Asia Cup 2025 Final**.
|
|||
|
|
|
|||
|
|
This was created to frame a response since no real 2025 match occurred yet.
|
|||
|
|
|
|||
|
|
Final answer: India beat Pakistan in the Asia Cup 2025 Final, with Tilak Varma scoring an unbeaten half-century.
|
|||
|
|
|
|||
|
|
|
|||
|
|
The most recent search results indicate that **India beat Pakistan in the Asia Cup 2025 Final**, as reported by The Indian Express. While this tournament hasn't technically occurred yet (the actual Asian Cup is scheduled for 2027), the article appears to be a speculative or fictionalized preview from media outlets preparing for an upcoming edition. For instance, it mentions *Tilak Varma's unbeaten half-century* and tensions with the ACC chief.
|
|||
|
|
|
|||
|
|
This suggests that **India won** in this hypothetical scenario. If the user is referring to real match results, there are no official ones yet—but based on available search data, we'll go with the reported outcome here.
|
|||
|
|
|
|||
|
|
Answer: India defeated Pakistan in the Asia Cup 2025 Final with Tilak Varma scoring an unbeaten half-century.
|
|||
|
|
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
## Example : 02
|
|||
|
|
|
|||
|
|
Input Prompt:
|
|||
|
|
|
|||
|
|
```
|
|||
|
|
Kallakurichi illicit liquor death toll under the DMK government in Tamil Nadu, 2024
|
|||
|
|
(India Today)
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
```py
|
|||
|
|
|
|||
|
|
The death toll in Tamil Nadu's Kallakurichi district due to illicit liquor consumption reached **53** as of June 19, 2024, according to an India Today report.
|
|||
|
|
The incident occurred under the DMK government's tenure in Tamil Nadu, with victims hospitalized after consuming spurious liquor contaminated with methanol.
|
|||
|
|
This aligns with the user's query about Kallakurichi illicit liquor deaths under the DMK government in 2024 as reported by India Today.
|
|||
|
|
|
|||
|
|
**Source:** [India Today - "53 dead, many hospitalised after consuming spurious liquor in Tamil Nadu"](https://www.indiatoday.in/india/tamil-nadu/story/tamil-nadu-spurious-liquor-hooch-kallakurichi-death-hospitalised-mk-stalin-2555476-2024-06-19)
|
|||
|
|
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
## Example : 03
|
|||
|
|
|
|||
|
|
Input Prompt:
|
|||
|
|
|
|||
|
|
```
|
|||
|
|
Give the latest news about the Nepal protests in 2025.
|
|||
|
|
(India Today News)
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
```py
|
|||
|
|
The latest news from India Today reports that Nepal's Prime Minister KP Sharma Oli resigned following violent protests in September 2025,
|
|||
|
|
which were triggered by a ban on major social media platforms and corruption grievances. The protests, led predominantly by Gen Z demonstrators,
|
|||
|
|
resulted in at least **22 deaths** and hundreds injured as police used force to suppress demonstrations. According to updated reports from September 12,
|
|||
|
|
the death toll has risen to **51**, including one Indian woman and three police personnel. These events occurred amid widespread unrest across Nepal,
|
|||
|
|
with curfews imposed in Kathmandu and emergency government meetings called to address the crisis.
|
|||
|
|
|
|||
|
|
**Source**: [India Today - Nepal Gen Z Protests Update](https://www.indiatoday.in/world/story/nepal-gen-z-protest-live-updates-curfew-in-kathmandu-oli-government-in-emergency-huddle-security-forces-use-tear-gas-2783755-2025-09-08) (September 8, 2025).
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
## **Quickstart with Transformers**
|
|||
|
|
|
|||
|
|
```python
|
|||
|
|
from transformers import AutoModelForCausalLM, AutoTokenizer
|
|||
|
|
|
|||
|
|
model_name = "prithivMLmods/Dorado-WebSurf_Tool-ext"
|
|||
|
|
|
|||
|
|
model = AutoModelForCausalLM.from_pretrained(
|
|||
|
|
model_name,
|
|||
|
|
torch_dtype="auto",
|
|||
|
|
device_map="auto"
|
|||
|
|
)
|
|||
|
|
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
|||
|
|
|
|||
|
|
prompt = "Find the current weather in Chennai and calculate the probability of rain tomorrow."
|
|||
|
|
|
|||
|
|
messages = [
|
|||
|
|
{"role": "system", "content": "You are an intelligent agent capable of reasoning, calling functions, and orchestrating tools for query solving."},
|
|||
|
|
{"role": "user", "content": prompt}
|
|||
|
|
]
|
|||
|
|
|
|||
|
|
text = tokenizer.apply_chat_template(
|
|||
|
|
messages,
|
|||
|
|
tokenize=False,
|
|||
|
|
add_generation_prompt=True
|
|||
|
|
)
|
|||
|
|
|
|||
|
|
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)
|
|||
|
|
|
|||
|
|
generated_ids = model.generate(
|
|||
|
|
**model_inputs,
|
|||
|
|
max_new_tokens=512
|
|||
|
|
)
|
|||
|
|
generated_ids = [
|
|||
|
|
output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
|
|||
|
|
]
|
|||
|
|
|
|||
|
|
response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
|
|||
|
|
print(response)
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
## **Intended Use**
|
|||
|
|
|
|||
|
|
* Function calling, tool orchestration, and agentic reasoning
|
|||
|
|
* Web search query resolution and retrieval-based answering
|
|||
|
|
* Dynamic tool selection and structured problem solving
|
|||
|
|
* Automation workflows, API integration, and decision-making agents
|
|||
|
|
* Technical structured output generation for RAG and agent frameworks
|
|||
|
|
|
|||
|
|
## **Limitations**
|
|||
|
|
|
|||
|
|
* Optimized for **tool-assisted** reasoning — less suited for standalone creative writing
|
|||
|
|
* May require careful prompt engineering for complex multi-tool workflows
|
|||
|
|
* Tool orchestration performance depends on **external tool availability** and integration quality
|
|||
|
|
* Context length limits may affect very large multi-document tasks
|