Add MiniMax as LLM provider (M2.7 + M2.5-highspeed)

README.md

@@ -7,7 +7,7 @@
 ### Control Your Computer Using LLMs
 
 Open Interface
-- Self-drives your computer by sending your requests to an LLM backend (GPT-4o, Gemini, etc) to figure out the required steps.
+- Self-drives your computer by sending your requests to an LLM backend (GPT-4o, Gemini, MiniMax, etc) to figure out the required steps.
 - Automatically executes these steps by simulating keyboard and mouse input.
 - Course-corrects by sending the LLM backend updated screenshots of the progress as needed.
 
@@ -172,6 +172,17 @@ Open Interface
 
 </details>
 
+<details>
+ <summary><b>Set up the MiniMax API key</b></summary>
+
+- Go to Settings -> Advanced Settings and select the MiniMax model you wish to use (MiniMax-M2.7 or MiniMax-M2.5-highspeed).
+- Get your MiniMax API key from https://platform.minimaxi.com/
+- Save the API key in Open Interface settings.
+- Save the settings and **restart the app**.
+- Alternatively, set the `MINIMAX_API_KEY` environment variable and the API key will be auto-detected.
+
+</details>
+
 <details>
  <summary><b>Optional: Setup a Custom LLM</b></summary>
 

app/models/factory.py

@@ -1,6 +1,7 @@
 from models.gpt4o import GPT4o
 from models.gpt4v import GPT4v
 from models.gpt5 import GPT5
+from models.minimax import MiniMax
 from models.openai_computer_use import OpenAIComputerUse
 from models.gemini import Gemini
 
@@ -19,6 +20,8 @@ def create_model(model_name, *args):
  return GPT4v(model_name, *args)
  elif model_name.startswith("gemini"):
  return Gemini(model_name, *args[1:])
+ elif model_name.startswith("MiniMax"):
+ return MiniMax(model_name, *args)
  else:
  # Llama/Llava models will work with the standard code I wrote for GPT4V without the assitant mode features of gpt4o
  return GPT4v(model_name, *args)

app/models/minimax.py

@@ -0,0 +1,87 @@
+import json
+import os
+from typing import Any
+
+from models.model import Model
+from utils.screen import Screen
+
+MINIMAX_BASE_URL = 'https://api.minimax.io/v1/'
+
+
+class MiniMax(Model):
+ """
+ MiniMax model integration via OpenAI-compatible API.
+ Supports MiniMax-M2.7 (latest) and MiniMax-M2.5-highspeed (204K context).
+ """
+
+ def __init__(self, model_name, base_url, api_key, context):
+ # Use MiniMax base URL unless user has explicitly set a custom one
+ if not base_url or base_url.rstrip('/') + '/' == 'https://api.openai.com/v1/':
+ base_url = MINIMAX_BASE_URL
+
+ # Prefer MINIMAX_API_KEY env var if no API key was provided
+ if not api_key:
+ api_key = os.environ.get('MINIMAX_API_KEY', '')
+
+ super().__init__(model_name, base_url, api_key, context)
+
+ if api_key:
+ os.environ['MINIMAX_API_KEY'] = api_key
+
+ def get_instructions_for_objective(self, original_user_request: str, step_num: int = 0) -> dict[str, Any]:
+ message: list[dict[str, Any]] = self.format_user_request_for_llm(original_user_request, step_num)
+ llm_response = self.send_message_to_llm(message)
+ json_instructions: dict[str, Any] = self.convert_llm_response_to_json_instructions(llm_response)
+ return json_instructions
+
+ def format_user_request_for_llm(self, original_user_request, step_num) -> list[dict[str, Any]]:
+ base64_img: str = Screen().get_screenshot_in_base64()
+
+ request_data: str = json.dumps({
+ 'original_user_request': original_user_request,
+ 'step_num': step_num
+ })
+
+ message = [
+ {'type': 'text', 'text': self.context + request_data},
+ {'type': 'image_url',
+ 'image_url': {
+ 'url': f'data:image/jpeg;base64,{base64_img}'
+ }
+ }
+ ]
+
+ return message
+
+ def send_message_to_llm(self, message):
+ response = self.client.chat.completions.create(
+ model=self.model_name,
+ messages=[
+ {
+ 'role': 'user',
+ 'content': message,
+ }
+ ],
+ max_tokens=800,
+ temperature=min(max(0.01, 0.7), 1.0),
+ )
+ return response
+
+ def convert_llm_response_to_json_instructions(self, llm_response) -> dict[str, Any]:
+ llm_response_data: str = llm_response.choices[0].message.content.strip()
+
+ # Strip MiniMax thinking tags if present (M2.5+ models may include <think>...</think> blocks)
+ if '<think>' in llm_response_data and '</think>' in llm_response_data:
+ think_end = llm_response_data.rfind('</think>')
+ llm_response_data = llm_response_data[think_end + len('</think>'):].strip()
+
+ start_index = llm_response_data.find('{')
+ end_index = llm_response_data.rfind('}')
+
+ try:
+ json_response = json.loads(llm_response_data[start_index:end_index + 1].strip())
+ except Exception as e:
+ print(f'Error while parsing JSON response - {e}')
+ json_response = {}
+
+ return json_response

app/ui.py

@@ -69,6 +69,11 @@ def create_widgets(self) -> None:
  ('Gemini gemini-3-flash-preview', 'gemini-3-flash-preview'),
  ]
 
+ minimax_models = [
+ ('MiniMax-M2.7 (1M Context)', 'MiniMax-M2.7'),
+ ('MiniMax-M2.5-highspeed (204K, Fast)', 'MiniMax-M2.5-highspeed'),
+ ]
+
  deprecated_models = [
  ('GPT-4o (Medium-Accurate, Medium-Fast)', 'gpt-4o'),
  ('GPT-4o-mini (Cheapest, Fastest)', 'gpt-4o-mini'),
@@ -93,6 +98,12 @@ def create_widgets(self) -> None:
  ttk.Radiobutton(radio_frame, text=text, value=value, variable=self.model_var, bootstyle="info").pack(
  anchor=ttk.W, pady=5)
 
+ ttk.Separator(radio_frame, orient='horizontal').pack(fill='x', pady=8)
+
+ for text, value in minimax_models:
+ ttk.Radiobutton(radio_frame, text=text, value=value, variable=self.model_var, bootstyle="info").pack(
+ anchor=ttk.W, pady=5)
+
  ttk.Separator(radio_frame, orient='horizontal').pack(fill='x', pady=10)
 
  self.deprecated_expanded = False
@@ -192,7 +203,7 @@ def __init__(self, parent):
 
  def create_widgets(self) -> None:
  # API Key Widgets
- label_api = ttk.Label(self, text='OpenAI/Gemini/LLM Model API Key:', bootstyle="info")
+ label_api = ttk.Label(self, text='OpenAI/Gemini/MiniMax/LLM Model API Key:', bootstyle="info")
  label_api.pack(pady=10)
  self.api_key_entry = ttk.Entry(self, width=30)
  self.api_key_entry.pack()

README.md

@@ -7,7 +7,7 @@
 ### Control Your Computer Using LLMs
 
 Open Interface
-- Self-drives your computer by sending your requests to an LLM backend (GPT-4o, Gemini, etc) to figure out the required steps.
+- Self-drives your computer by sending your requests to an LLM backend (GPT-4o, Gemini, MiniMax, etc) to figure out the required steps.
 - Automatically executes these steps by simulating keyboard and mouse input.
 - Course-corrects by sending the LLM backend updated screenshots of the progress as needed.
 
@@ -172,6 +172,17 @@ Open Interface
 
 </details>
 
+<details>
+ <summary><b>Set up the MiniMax API key</b></summary>
+
+- Go to Settings -> Advanced Settings and select the MiniMax model you wish to use (MiniMax-M2.7 or MiniMax-M2.5-highspeed).
+- Get your MiniMax API key from https://platform.minimaxi.com/
+- Save the API key in Open Interface settings.
+- Save the settings and **restart the app**.
+- Alternatively, set the `MINIMAX_API_KEY` environment variable and the API key will be auto-detected.
+
+</details>
+
 <details>
  <summary><b>Optional: Setup a Custom LLM</b></summary>
 

app/models/factory.py

@@ -1,6 +1,7 @@
 from models.gpt4o import GPT4o
 from models.gpt4v import GPT4v
 from models.gpt5 import GPT5
+from models.minimax import MiniMax
 from models.openai_computer_use import OpenAIComputerUse
 from models.gemini import Gemini
 
@@ -19,6 +20,8 @@ def create_model(model_name, *args):
  return GPT4v(model_name, *args)
  elif model_name.startswith("gemini"):
  return Gemini(model_name, *args[1:])
+ elif model_name.startswith("MiniMax"):
+ return MiniMax(model_name, *args)
  else:
  # Llama/Llava models will work with the standard code I wrote for GPT4V without the assitant mode features of gpt4o
  return GPT4v(model_name, *args)

app/models/minimax.py

@@ -0,0 +1,87 @@
+import json
+import os
+from typing import Any
+
+from models.model import Model
+from utils.screen import Screen
+
+MINIMAX_BASE_URL = 'https://api.minimax.io/v1/'
+
+
+class MiniMax(Model):
+ """
+ MiniMax model integration via OpenAI-compatible API.
+ Supports MiniMax-M2.7 (latest) and MiniMax-M2.5-highspeed (204K context).
+ """
+
+ def __init__(self, model_name, base_url, api_key, context):
+ # Use MiniMax base URL unless user has explicitly set a custom one
+ if not base_url or base_url.rstrip('/') + '/' == 'https://api.openai.com/v1/':
+ base_url = MINIMAX_BASE_URL
+
+ # Prefer MINIMAX_API_KEY env var if no API key was provided
+ if not api_key:
+ api_key = os.environ.get('MINIMAX_API_KEY', '')
+
+ super().__init__(model_name, base_url, api_key, context)
+
+ if api_key:
+ os.environ['MINIMAX_API_KEY'] = api_key
+
+ def get_instructions_for_objective(self, original_user_request: str, step_num: int = 0) -> dict[str, Any]:
+ message: list[dict[str, Any]] = self.format_user_request_for_llm(original_user_request, step_num)
+ llm_response = self.send_message_to_llm(message)
+ json_instructions: dict[str, Any] = self.convert_llm_response_to_json_instructions(llm_response)
+ return json_instructions
+
+ def format_user_request_for_llm(self, original_user_request, step_num) -> list[dict[str, Any]]:
+ base64_img: str = Screen().get_screenshot_in_base64()
+
+ request_data: str = json.dumps({
+ 'original_user_request': original_user_request,
+ 'step_num': step_num
+ })
+
+ message = [
+ {'type': 'text', 'text': self.context + request_data},
+ {'type': 'image_url',
+ 'image_url': {
+ 'url': f'data:image/jpeg;base64,{base64_img}'
+ }
+ }
+ ]
+
+ return message
+
+ def send_message_to_llm(self, message):
+ response = self.client.chat.completions.create(
+ model=self.model_name,
+ messages=[
+ {
+ 'role': 'user',
+ 'content': message,
+ }
+ ],
+ max_tokens=800,
+ temperature=min(max(0.01, 0.7), 1.0),
+ )
+ return response
+
+ def convert_llm_response_to_json_instructions(self, llm_response) -> dict[str, Any]:
+ llm_response_data: str = llm_response.choices[0].message.content.strip()
+
+ # Strip MiniMax thinking tags if present (M2.5+ models may include <think>...</think> blocks)
+ if '<think>' in llm_response_data and '</think>' in llm_response_data:
+ think_end = llm_response_data.rfind('</think>')
+ llm_response_data = llm_response_data[think_end + len('</think>'):].strip()
+
+ start_index = llm_response_data.find('{')
+ end_index = llm_response_data.rfind('}')
+
+ try:
+ json_response = json.loads(llm_response_data[start_index:end_index + 1].strip())
+ except Exception as e:
+ print(f'Error while parsing JSON response - {e}')
+ json_response = {}
+
+ return json_response

app/ui.py

@@ -69,6 +69,11 @@ def create_widgets(self) -> None:
  ('Gemini gemini-3-flash-preview', 'gemini-3-flash-preview'),
  ]
 
+ minimax_models = [
+ ('MiniMax-M2.7 (1M Context)', 'MiniMax-M2.7'),
+ ('MiniMax-M2.5-highspeed (204K, Fast)', 'MiniMax-M2.5-highspeed'),
+ ]
+
  deprecated_models = [
  ('GPT-4o (Medium-Accurate, Medium-Fast)', 'gpt-4o'),
  ('GPT-4o-mini (Cheapest, Fastest)', 'gpt-4o-mini'),
@@ -93,6 +98,12 @@ def create_widgets(self) -> None:
  ttk.Radiobutton(radio_frame, text=text, value=value, variable=self.model_var, bootstyle="info").pack(
  anchor=ttk.W, pady=5)
 
+ ttk.Separator(radio_frame, orient='horizontal').pack(fill='x', pady=8)
+
+ for text, value in minimax_models:
+ ttk.Radiobutton(radio_frame, text=text, value=value, variable=self.model_var, bootstyle="info").pack(
+ anchor=ttk.W, pady=5)
+
  ttk.Separator(radio_frame, orient='horizontal').pack(fill='x', pady=10)
 
  self.deprecated_expanded = False
@@ -192,7 +203,7 @@ def __init__(self, parent):
 
  def create_widgets(self) -> None:
  # API Key Widgets
- label_api = ttk.Label(self, text='OpenAI/Gemini/LLM Model API Key:', bootstyle="info")
+ label_api = ttk.Label(self, text='OpenAI/Gemini/MiniMax/LLM Model API Key:', bootstyle="info")
  label_api.pack(pady=10)
  self.api_key_entry = ttk.Entry(self, width=30)
  self.api_key_entry.pack()

README.md

@@ -7,7 +7,7 @@
 ### Control Your Computer Using LLMs
 
 Open Interface
-- Self-drives your computer by sending your requests to an LLM backend (GPT-4o, Gemini, etc) to figure out the required steps.
+- Self-drives your computer by sending your requests to an LLM backend (GPT-4o, Gemini, MiniMax, etc) to figure out the required steps.
 - Automatically executes these steps by simulating keyboard and mouse input.
 - Course-corrects by sending the LLM backend updated screenshots of the progress as needed.
 
@@ -172,6 +172,17 @@ Open Interface
 
 </details>
 
+<details>
+ <summary><b>Set up the MiniMax API key</b></summary>
+
+- Go to Settings -> Advanced Settings and select the MiniMax model you wish to use (MiniMax-M2.7 or MiniMax-M2.5-highspeed).
+- Get your MiniMax API key from https://platform.minimaxi.com/
+- Save the API key in Open Interface settings.
+- Save the settings and **restart the app**.
+- Alternatively, set the `MINIMAX_API_KEY` environment variable and the API key will be auto-detected.
+
+</details>
+
 <details>
  <summary><b>Optional: Setup a Custom LLM</b></summary>
 

app/models/factory.py

@@ -1,6 +1,7 @@
 from models.gpt4o import GPT4o
 from models.gpt4v import GPT4v
 from models.gpt5 import GPT5
+from models.minimax import MiniMax
 from models.openai_computer_use import OpenAIComputerUse
 from models.gemini import Gemini
 
@@ -19,6 +20,8 @@ def create_model(model_name, *args):
  return GPT4v(model_name, *args)
  elif model_name.startswith("gemini"):
  return Gemini(model_name, *args[1:])
+ elif model_name.startswith("MiniMax"):
+ return MiniMax(model_name, *args)
  else:
  # Llama/Llava models will work with the standard code I wrote for GPT4V without the assitant mode features of gpt4o
  return GPT4v(model_name, *args)

app/models/minimax.py

@@ -0,0 +1,87 @@
+import json
+import os
+from typing import Any
+
+from models.model import Model
+from utils.screen import Screen
+
+MINIMAX_BASE_URL = 'https://api.minimax.io/v1/'
+
+
+class MiniMax(Model):
+ """
+ MiniMax model integration via OpenAI-compatible API.
+ Supports MiniMax-M2.7 (latest) and MiniMax-M2.5-highspeed (204K context).
+ """
+
+ def __init__(self, model_name, base_url, api_key, context):
+ # Use MiniMax base URL unless user has explicitly set a custom one
+ if not base_url or base_url.rstrip('/') + '/' == 'https://api.openai.com/v1/':
+ base_url = MINIMAX_BASE_URL
+
+ # Prefer MINIMAX_API_KEY env var if no API key was provided
+ if not api_key:
+ api_key = os.environ.get('MINIMAX_API_KEY', '')
+
+ super().__init__(model_name, base_url, api_key, context)
+
+ if api_key:
+ os.environ['MINIMAX_API_KEY'] = api_key
+
+ def get_instructions_for_objective(self, original_user_request: str, step_num: int = 0) -> dict[str, Any]:
+ message: list[dict[str, Any]] = self.format_user_request_for_llm(original_user_request, step_num)
+ llm_response = self.send_message_to_llm(message)
+ json_instructions: dict[str, Any] = self.convert_llm_response_to_json_instructions(llm_response)
+ return json_instructions
+
+ def format_user_request_for_llm(self, original_user_request, step_num) -> list[dict[str, Any]]:
+ base64_img: str = Screen().get_screenshot_in_base64()
+
+ request_data: str = json.dumps({
+ 'original_user_request': original_user_request,
+ 'step_num': step_num
+ })
+
+ message = [
+ {'type': 'text', 'text': self.context + request_data},
+ {'type': 'image_url',
+ 'image_url': {
+ 'url': f'data:image/jpeg;base64,{base64_img}'
+ }
+ }
+ ]
+
+ return message
+
+ def send_message_to_llm(self, message):
+ response = self.client.chat.completions.create(
+ model=self.model_name,
+ messages=[
+ {
+ 'role': 'user',
+ 'content': message,
+ }
+ ],
+ max_tokens=800,
+ temperature=min(max(0.01, 0.7), 1.0),
+ )
+ return response
+
+ def convert_llm_response_to_json_instructions(self, llm_response) -> dict[str, Any]:
+ llm_response_data: str = llm_response.choices[0].message.content.strip()
+
+ # Strip MiniMax thinking tags if present (M2.5+ models may include <think>...</think> blocks)
+ if '<think>' in llm_response_data and '</think>' in llm_response_data:
+ think_end = llm_response_data.rfind('</think>')
+ llm_response_data = llm_response_data[think_end + len('</think>'):].strip()
+
+ start_index = llm_response_data.find('{')
+ end_index = llm_response_data.rfind('}')
+
+ try:
+ json_response = json.loads(llm_response_data[start_index:end_index + 1].strip())
+ except Exception as e:
+ print(f'Error while parsing JSON response - {e}')
+ json_response = {}
+
+ return json_response

app/ui.py

@@ -69,6 +69,11 @@ def create_widgets(self) -> None:
  ('Gemini gemini-3-flash-preview', 'gemini-3-flash-preview'),
  ]
 
+ minimax_models = [
+ ('MiniMax-M2.7 (1M Context)', 'MiniMax-M2.7'),
+ ('MiniMax-M2.5-highspeed (204K, Fast)', 'MiniMax-M2.5-highspeed'),
+ ]
+
  deprecated_models = [
  ('GPT-4o (Medium-Accurate, Medium-Fast)', 'gpt-4o'),
  ('GPT-4o-mini (Cheapest, Fastest)', 'gpt-4o-mini'),
@@ -93,6 +98,12 @@ def create_widgets(self) -> None:
  ttk.Radiobutton(radio_frame, text=text, value=value, variable=self.model_var, bootstyle="info").pack(
  anchor=ttk.W, pady=5)
 
+ ttk.Separator(radio_frame, orient='horizontal').pack(fill='x', pady=8)
+
+ for text, value in minimax_models:
+ ttk.Radiobutton(radio_frame, text=text, value=value, variable=self.model_var, bootstyle="info").pack(
+ anchor=ttk.W, pady=5)
+
  ttk.Separator(radio_frame, orient='horizontal').pack(fill='x', pady=10)
 
  self.deprecated_expanded = False
@@ -192,7 +203,7 @@ def __init__(self, parent):
 
  def create_widgets(self) -> None:
  # API Key Widgets
- label_api = ttk.Label(self, text='OpenAI/Gemini/LLM Model API Key:', bootstyle="info")
+ label_api = ttk.Label(self, text='OpenAI/Gemini/MiniMax/LLM Model API Key:', bootstyle="info")
  label_api.pack(pady=10)
  self.api_key_entry = ttk.Entry(self, width=30)
  self.api_key_entry.pack()

README.md

@@ -7,7 +7,7 @@
 ### Control Your Computer Using LLMs
 
 Open Interface
-- Self-drives your computer by sending your requests to an LLM backend (GPT-4o, Gemini, etc) to figure out the required steps.
+- Self-drives your computer by sending your requests to an LLM backend (GPT-4o, Gemini, MiniMax, etc) to figure out the required steps.
 - Automatically executes these steps by simulating keyboard and mouse input.
 - Course-corrects by sending the LLM backend updated screenshots of the progress as needed.
 
@@ -172,6 +172,17 @@ Open Interface
 
 </details>
 
+<details>
+ <summary><b>Set up the MiniMax API key</b></summary>
+
+- Go to Settings -> Advanced Settings and select the MiniMax model you wish to use (MiniMax-M2.7 or MiniMax-M2.5-highspeed).
+- Get your MiniMax API key from https://platform.minimaxi.com/
+- Save the API key in Open Interface settings.
+- Save the settings and **restart the app**.
+- Alternatively, set the `MINIMAX_API_KEY` environment variable and the API key will be auto-detected.
+
+</details>
+
 <details>
  <summary><b>Optional: Setup a Custom LLM</b></summary>
 

app/models/factory.py

@@ -1,6 +1,7 @@
 from models.gpt4o import GPT4o
 from models.gpt4v import GPT4v
 from models.gpt5 import GPT5
+from models.minimax import MiniMax
 from models.openai_computer_use import OpenAIComputerUse
 from models.gemini import Gemini
 
@@ -19,6 +20,8 @@ def create_model(model_name, *args):
  return GPT4v(model_name, *args)
  elif model_name.startswith("gemini"):
  return Gemini(model_name, *args[1:])
+ elif model_name.startswith("MiniMax"):
+ return MiniMax(model_name, *args)
  else:
  # Llama/Llava models will work with the standard code I wrote for GPT4V without the assitant mode features of gpt4o
  return GPT4v(model_name, *args)

app/models/minimax.py

@@ -0,0 +1,87 @@
+import json
+import os
+from typing import Any
+
+from models.model import Model
+from utils.screen import Screen
+
+MINIMAX_BASE_URL = 'https://api.minimax.io/v1/'
+
+
+class MiniMax(Model):
+ """
+ MiniMax model integration via OpenAI-compatible API.
+ Supports MiniMax-M2.7 (latest) and MiniMax-M2.5-highspeed (204K context).
+ """
+
+ def __init__(self, model_name, base_url, api_key, context):
+ # Use MiniMax base URL unless user has explicitly set a custom one
+ if not base_url or base_url.rstrip('/') + '/' == 'https://api.openai.com/v1/':
+ base_url = MINIMAX_BASE_URL
+
+ # Prefer MINIMAX_API_KEY env var if no API key was provided
+ if not api_key:
+ api_key = os.environ.get('MINIMAX_API_KEY', '')
+
+ super().__init__(model_name, base_url, api_key, context)
+
+ if api_key:
+ os.environ['MINIMAX_API_KEY'] = api_key
+
+ def get_instructions_for_objective(self, original_user_request: str, step_num: int = 0) -> dict[str, Any]:
+ message: list[dict[str, Any]] = self.format_user_request_for_llm(original_user_request, step_num)
+ llm_response = self.send_message_to_llm(message)
+ json_instructions: dict[str, Any] = self.convert_llm_response_to_json_instructions(llm_response)
+ return json_instructions
+
+ def format_user_request_for_llm(self, original_user_request, step_num) -> list[dict[str, Any]]:
+ base64_img: str = Screen().get_screenshot_in_base64()
+
+ request_data: str = json.dumps({
+ 'original_user_request': original_user_request,
+ 'step_num': step_num
+ })
+
+ message = [
+ {'type': 'text', 'text': self.context + request_data},
+ {'type': 'image_url',
+ 'image_url': {
+ 'url': f'data:image/jpeg;base64,{base64_img}'
+ }
+ }
+ ]
+
+ return message
+
+ def send_message_to_llm(self, message):
+ response = self.client.chat.completions.create(
+ model=self.model_name,
+ messages=[
+ {
+ 'role': 'user',
+ 'content': message,
+ }
+ ],
+ max_tokens=800,
+ temperature=min(max(0.01, 0.7), 1.0),
+ )
+ return response
+
+ def convert_llm_response_to_json_instructions(self, llm_response) -> dict[str, Any]:
+ llm_response_data: str = llm_response.choices[0].message.content.strip()
+
+ # Strip MiniMax thinking tags if present (M2.5+ models may include <think>...</think> blocks)
+ if '<think>' in llm_response_data and '</think>' in llm_response_data:
+ think_end = llm_response_data.rfind('</think>')
+ llm_response_data = llm_response_data[think_end + len('</think>'):].strip()
+
+ start_index = llm_response_data.find('{')
+ end_index = llm_response_data.rfind('}')
+
+ try:
+ json_response = json.loads(llm_response_data[start_index:end_index + 1].strip())
+ except Exception as e:
+ print(f'Error while parsing JSON response - {e}')
+ json_response = {}
+
+ return json_response

app/ui.py

@@ -69,6 +69,11 @@ def create_widgets(self) -> None:
  ('Gemini gemini-3-flash-preview', 'gemini-3-flash-preview'),
  ]
 
+ minimax_models = [
+ ('MiniMax-M2.7 (1M Context)', 'MiniMax-M2.7'),
+ ('MiniMax-M2.5-highspeed (204K, Fast)', 'MiniMax-M2.5-highspeed'),
+ ]
+
  deprecated_models = [
  ('GPT-4o (Medium-Accurate, Medium-Fast)', 'gpt-4o'),
  ('GPT-4o-mini (Cheapest, Fastest)', 'gpt-4o-mini'),
@@ -93,6 +98,12 @@ def create_widgets(self) -> None:
  ttk.Radiobutton(radio_frame, text=text, value=value, variable=self.model_var, bootstyle="info").pack(
  anchor=ttk.W, pady=5)
 
+ ttk.Separator(radio_frame, orient='horizontal').pack(fill='x', pady=8)
+
+ for text, value in minimax_models:
+ ttk.Radiobutton(radio_frame, text=text, value=value, variable=self.model_var, bootstyle="info").pack(
+ anchor=ttk.W, pady=5)
+
  ttk.Separator(radio_frame, orient='horizontal').pack(fill='x', pady=10)
 
  self.deprecated_expanded = False
@@ -192,7 +203,7 @@ def __init__(self, parent):
 
  def create_widgets(self) -> None:
  # API Key Widgets
- label_api = ttk.Label(self, text='OpenAI/Gemini/LLM Model API Key:', bootstyle="info")
+ label_api = ttk.Label(self, text='OpenAI/Gemini/MiniMax/LLM Model API Key:', bootstyle="info")
  label_api.pack(pady=10)
  self.api_key_entry = ttk.Entry(self, width=30)
  self.api_key_entry.pack()

README.md

@@ -7,7 +7,7 @@
 ### Control Your Computer Using LLMs
 
 Open Interface
-- Self-drives your computer by sending your requests to an LLM backend (GPT-4o, Gemini, etc) to figure out the required steps.
+- Self-drives your computer by sending your requests to an LLM backend (GPT-4o, Gemini, MiniMax, etc) to figure out the required steps.
 - Automatically executes these steps by simulating keyboard and mouse input.
 - Course-corrects by sending the LLM backend updated screenshots of the progress as needed.
 
@@ -172,6 +172,17 @@ Open Interface
 
 </details>
 
+<details>
+ <summary><b>Set up the MiniMax API key</b></summary>
+
+- Go to Settings -> Advanced Settings and select the MiniMax model you wish to use (MiniMax-M2.7 or MiniMax-M2.5-highspeed).
+- Get your MiniMax API key from https://platform.minimaxi.com/
+- Save the API key in Open Interface settings.
+- Save the settings and **restart the app**.
+- Alternatively, set the `MINIMAX_API_KEY` environment variable and the API key will be auto-detected.
+
+</details>
+
 <details>
  <summary><b>Optional: Setup a Custom LLM</b></summary>
 

app/models/factory.py

@@ -1,6 +1,7 @@
 from models.gpt4o import GPT4o
 from models.gpt4v import GPT4v
 from models.gpt5 import GPT5
+from models.minimax import MiniMax
 from models.openai_computer_use import OpenAIComputerUse
 from models.gemini import Gemini
 
@@ -19,6 +20,8 @@ def create_model(model_name, *args):
  return GPT4v(model_name, *args)
  elif model_name.startswith("gemini"):
  return Gemini(model_name, *args[1:])
+ elif model_name.startswith("MiniMax"):
+ return MiniMax(model_name, *args)
  else:
  # Llama/Llava models will work with the standard code I wrote for GPT4V without the assitant mode features of gpt4o
  return GPT4v(model_name, *args)

app/models/minimax.py

@@ -0,0 +1,87 @@
+import json
+import os
+from typing import Any
+
+from models.model import Model
+from utils.screen import Screen
+
+MINIMAX_BASE_URL = 'https://api.minimax.io/v1/'
+
+
+class MiniMax(Model):
+ """
+ MiniMax model integration via OpenAI-compatible API.
+ Supports MiniMax-M2.7 (latest) and MiniMax-M2.5-highspeed (204K context).
+ """
+
+ def __init__(self, model_name, base_url, api_key, context):
+ # Use MiniMax base URL unless user has explicitly set a custom one
+ if not base_url or base_url.rstrip('/') + '/' == 'https://api.openai.com/v1/':
+ base_url = MINIMAX_BASE_URL
+
+ # Prefer MINIMAX_API_KEY env var if no API key was provided
+ if not api_key:
+ api_key = os.environ.get('MINIMAX_API_KEY', '')
+
+ super().__init__(model_name, base_url, api_key, context)
+
+ if api_key:
+ os.environ['MINIMAX_API_KEY'] = api_key
+
+ def get_instructions_for_objective(self, original_user_request: str, step_num: int = 0) -> dict[str, Any]:
+ message: list[dict[str, Any]] = self.format_user_request_for_llm(original_user_request, step_num)
+ llm_response = self.send_message_to_llm(message)
+ json_instructions: dict[str, Any] = self.convert_llm_response_to_json_instructions(llm_response)
+ return json_instructions
+
+ def format_user_request_for_llm(self, original_user_request, step_num) -> list[dict[str, Any]]:
+ base64_img: str = Screen().get_screenshot_in_base64()
+
+ request_data: str = json.dumps({
+ 'original_user_request': original_user_request,
+ 'step_num': step_num
+ })
+
+ message = [
+ {'type': 'text', 'text': self.context + request_data},
+ {'type': 'image_url',
+ 'image_url': {
+ 'url': f'data:image/jpeg;base64,{base64_img}'
+ }
+ }
+ ]
+
+ return message
+
+ def send_message_to_llm(self, message):
+ response = self.client.chat.completions.create(
+ model=self.model_name,
+ messages=[
+ {
+ 'role': 'user',
+ 'content': message,
+ }
+ ],
+ max_tokens=800,
+ temperature=min(max(0.01, 0.7), 1.0),
+ )
+ return response
+
+ def convert_llm_response_to_json_instructions(self, llm_response) -> dict[str, Any]:
+ llm_response_data: str = llm_response.choices[0].message.content.strip()
+
+ # Strip MiniMax thinking tags if present (M2.5+ models may include <think>...</think> blocks)
+ if '<think>' in llm_response_data and '</think>' in llm_response_data:
+ think_end = llm_response_data.rfind('</think>')
+ llm_response_data = llm_response_data[think_end + len('</think>'):].strip()
+
+ start_index = llm_response_data.find('{')
+ end_index = llm_response_data.rfind('}')
+
+ try:
+ json_response = json.loads(llm_response_data[start_index:end_index + 1].strip())
+ except Exception as e:
+ print(f'Error while parsing JSON response - {e}')
+ json_response = {}
+
+ return json_response

app/ui.py

@@ -69,6 +69,11 @@ def create_widgets(self) -> None:
  ('Gemini gemini-3-flash-preview', 'gemini-3-flash-preview'),
  ]
 
+ minimax_models = [
+ ('MiniMax-M2.7 (1M Context)', 'MiniMax-M2.7'),
+ ('MiniMax-M2.5-highspeed (204K, Fast)', 'MiniMax-M2.5-highspeed'),
+ ]
+
  deprecated_models = [
  ('GPT-4o (Medium-Accurate, Medium-Fast)', 'gpt-4o'),
  ('GPT-4o-mini (Cheapest, Fastest)', 'gpt-4o-mini'),
@@ -93,6 +98,12 @@ def create_widgets(self) -> None:
  ttk.Radiobutton(radio_frame, text=text, value=value, variable=self.model_var, bootstyle="info").pack(
  anchor=ttk.W, pady=5)
 
+ ttk.Separator(radio_frame, orient='horizontal').pack(fill='x', pady=8)
+
+ for text, value in minimax_models:
+ ttk.Radiobutton(radio_frame, text=text, value=value, variable=self.model_var, bootstyle="info").pack(
+ anchor=ttk.W, pady=5)
+
  ttk.Separator(radio_frame, orient='horizontal').pack(fill='x', pady=10)
 
  self.deprecated_expanded = False
@@ -192,7 +203,7 @@ def __init__(self, parent):
 
  def create_widgets(self) -> None:
  # API Key Widgets
- label_api = ttk.Label(self, text='OpenAI/Gemini/LLM Model API Key:', bootstyle="info")
+ label_api = ttk.Label(self, text='OpenAI/Gemini/MiniMax/LLM Model API Key:', bootstyle="info")
  label_api.pack(pady=10)
  self.api_key_entry = ttk.Entry(self, width=30)
  self.api_key_entry.pack()

README.md

@@ -7,7 +7,7 @@
 ### Control Your Computer Using LLMs
 
 Open Interface
-- Self-drives your computer by sending your requests to an LLM backend (GPT-4o, Gemini, etc) to figure out the required steps.
+- Self-drives your computer by sending your requests to an LLM backend (GPT-4o, Gemini, MiniMax, etc) to figure out the required steps.
 - Automatically executes these steps by simulating keyboard and mouse input.
 - Course-corrects by sending the LLM backend updated screenshots of the progress as needed.
 
@@ -172,6 +172,17 @@ Open Interface
 
 </details>
 
+<details>
+ <summary><b>Set up the MiniMax API key</b></summary>
+
+- Go to Settings -> Advanced Settings and select the MiniMax model you wish to use (MiniMax-M2.7 or MiniMax-M2.5-highspeed).
+- Get your MiniMax API key from https://platform.minimaxi.com/
+- Save the API key in Open Interface settings.
+- Save the settings and **restart the app**.
+- Alternatively, set the `MINIMAX_API_KEY` environment variable and the API key will be auto-detected.
+
+</details>
+
 <details>
  <summary><b>Optional: Setup a Custom LLM</b></summary>
 

app/models/factory.py

@@ -1,6 +1,7 @@
 from models.gpt4o import GPT4o
 from models.gpt4v import GPT4v
 from models.gpt5 import GPT5
+from models.minimax import MiniMax
 from models.openai_computer_use import OpenAIComputerUse
 from models.gemini import Gemini
 
@@ -19,6 +20,8 @@ def create_model(model_name, *args):
  return GPT4v(model_name, *args)
  elif model_name.startswith("gemini"):
  return Gemini(model_name, *args[1:])
+ elif model_name.startswith("MiniMax"):
+ return MiniMax(model_name, *args)
  else:
  # Llama/Llava models will work with the standard code I wrote for GPT4V without the assitant mode features of gpt4o
  return GPT4v(model_name, *args)

app/models/minimax.py

@@ -0,0 +1,87 @@
+import json
+import os
+from typing import Any
+
+from models.model import Model
+from utils.screen import Screen
+
+MINIMAX_BASE_URL = 'https://api.minimax.io/v1/'
+
+
+class MiniMax(Model):
+ """
+ MiniMax model integration via OpenAI-compatible API.
+ Supports MiniMax-M2.7 (latest) and MiniMax-M2.5-highspeed (204K context).
+ """
+
+ def __init__(self, model_name, base_url, api_key, context):
+ # Use MiniMax base URL unless user has explicitly set a custom one
+ if not base_url or base_url.rstrip('/') + '/' == 'https://api.openai.com/v1/':
+ base_url = MINIMAX_BASE_URL
+
+ # Prefer MINIMAX_API_KEY env var if no API key was provided
+ if not api_key:
+ api_key = os.environ.get('MINIMAX_API_KEY', '')
+
+ super().__init__(model_name, base_url, api_key, context)
+
+ if api_key:
+ os.environ['MINIMAX_API_KEY'] = api_key
+
+ def get_instructions_for_objective(self, original_user_request: str, step_num: int = 0) -> dict[str, Any]:
+ message: list[dict[str, Any]] = self.format_user_request_for_llm(original_user_request, step_num)
+ llm_response = self.send_message_to_llm(message)
+ json_instructions: dict[str, Any] = self.convert_llm_response_to_json_instructions(llm_response)
+ return json_instructions
+
+ def format_user_request_for_llm(self, original_user_request, step_num) -> list[dict[str, Any]]:
+ base64_img: str = Screen().get_screenshot_in_base64()
+
+ request_data: str = json.dumps({
+ 'original_user_request': original_user_request,
+ 'step_num': step_num
+ })
+
+ message = [
+ {'type': 'text', 'text': self.context + request_data},
+ {'type': 'image_url',
+ 'image_url': {
+ 'url': f'data:image/jpeg;base64,{base64_img}'
+ }
+ }
+ ]
+
+ return message
+
+ def send_message_to_llm(self, message):
+ response = self.client.chat.completions.create(
+ model=self.model_name,
+ messages=[
+ {
+ 'role': 'user',
+ 'content': message,
+ }
+ ],
+ max_tokens=800,
+ temperature=min(max(0.01, 0.7), 1.0),
+ )
+ return response
+
+ def convert_llm_response_to_json_instructions(self, llm_response) -> dict[str, Any]:
+ llm_response_data: str = llm_response.choices[0].message.content.strip()
+
+ # Strip MiniMax thinking tags if present (M2.5+ models may include <think>...</think> blocks)
+ if '<think>' in llm_response_data and '</think>' in llm_response_data:
+ think_end = llm_response_data.rfind('</think>')
+ llm_response_data = llm_response_data[think_end + len('</think>'):].strip()
+
+ start_index = llm_response_data.find('{')
+ end_index = llm_response_data.rfind('}')
+
+ try:
+ json_response = json.loads(llm_response_data[start_index:end_index + 1].strip())
+ except Exception as e:
+ print(f'Error while parsing JSON response - {e}')
+ json_response = {}
+
+ return json_response

app/ui.py

@@ -69,6 +69,11 @@ def create_widgets(self) -> None:
  ('Gemini gemini-3-flash-preview', 'gemini-3-flash-preview'),
  ]
 
+ minimax_models = [
+ ('MiniMax-M2.7 (1M Context)', 'MiniMax-M2.7'),
+ ('MiniMax-M2.5-highspeed (204K, Fast)', 'MiniMax-M2.5-highspeed'),
+ ]
+
  deprecated_models = [
  ('GPT-4o (Medium-Accurate, Medium-Fast)', 'gpt-4o'),
  ('GPT-4o-mini (Cheapest, Fastest)', 'gpt-4o-mini'),
@@ -93,6 +98,12 @@ def create_widgets(self) -> None:
  ttk.Radiobutton(radio_frame, text=text, value=value, variable=self.model_var, bootstyle="info").pack(
  anchor=ttk.W, pady=5)
 
+ ttk.Separator(radio_frame, orient='horizontal').pack(fill='x', pady=8)
+
+ for text, value in minimax_models:
+ ttk.Radiobutton(radio_frame, text=text, value=value, variable=self.model_var, bootstyle="info").pack(
+ anchor=ttk.W, pady=5)
+
  ttk.Separator(radio_frame, orient='horizontal').pack(fill='x', pady=10)
 
  self.deprecated_expanded = False
@@ -192,7 +203,7 @@ def __init__(self, parent):
 
  def create_widgets(self) -> None:
  # API Key Widgets
- label_api = ttk.Label(self, text='OpenAI/Gemini/LLM Model API Key:', bootstyle="info")
+ label_api = ttk.Label(self, text='OpenAI/Gemini/MiniMax/LLM Model API Key:', bootstyle="info")
  label_api.pack(pady=10)
  self.api_key_entry = ttk.Entry(self, width=30)
  self.api_key_entry.pack()

README.md

@@ -7,7 +7,7 @@
 ### Control Your Computer Using LLMs
 
 Open Interface
-- Self-drives your computer by sending your requests to an LLM backend (GPT-4o, Gemini, etc) to figure out the required steps.
+- Self-drives your computer by sending your requests to an LLM backend (GPT-4o, Gemini, MiniMax, etc) to figure out the required steps.
 - Automatically executes these steps by simulating keyboard and mouse input.
 - Course-corrects by sending the LLM backend updated screenshots of the progress as needed.
 
@@ -172,6 +172,17 @@ Open Interface
 
 </details>
 
+<details>
+ <summary><b>Set up the MiniMax API key</b></summary>
+
+- Go to Settings -> Advanced Settings and select the MiniMax model you wish to use (MiniMax-M2.7 or MiniMax-M2.5-highspeed).
+- Get your MiniMax API key from https://platform.minimaxi.com/
+- Save the API key in Open Interface settings.
+- Save the settings and **restart the app**.
+- Alternatively, set the `MINIMAX_API_KEY` environment variable and the API key will be auto-detected.
+
+</details>
+
 <details>
  <summary><b>Optional: Setup a Custom LLM</b></summary>
 

app/models/factory.py

@@ -1,6 +1,7 @@
 from models.gpt4o import GPT4o
 from models.gpt4v import GPT4v
 from models.gpt5 import GPT5
+from models.minimax import MiniMax
 from models.openai_computer_use import OpenAIComputerUse
 from models.gemini import Gemini
 
@@ -19,6 +20,8 @@ def create_model(model_name, *args):
  return GPT4v(model_name, *args)
  elif model_name.startswith("gemini"):
  return Gemini(model_name, *args[1:])
+ elif model_name.startswith("MiniMax"):
+ return MiniMax(model_name, *args)
  else:
  # Llama/Llava models will work with the standard code I wrote for GPT4V without the assitant mode features of gpt4o
  return GPT4v(model_name, *args)

app/models/minimax.py

@@ -0,0 +1,87 @@
+import json
+import os
+from typing import Any
+
+from models.model import Model
+from utils.screen import Screen
+
+MINIMAX_BASE_URL = 'https://api.minimax.io/v1/'
+
+
+class MiniMax(Model):
+ """
+ MiniMax model integration via OpenAI-compatible API.
+ Supports MiniMax-M2.7 (latest) and MiniMax-M2.5-highspeed (204K context).
+ """
+
+ def __init__(self, model_name, base_url, api_key, context):
+ # Use MiniMax base URL unless user has explicitly set a custom one
+ if not base_url or base_url.rstrip('/') + '/' == 'https://api.openai.com/v1/':
+ base_url = MINIMAX_BASE_URL
+
+ # Prefer MINIMAX_API_KEY env var if no API key was provided
+ if not api_key:
+ api_key = os.environ.get('MINIMAX_API_KEY', '')
+
+ super().__init__(model_name, base_url, api_key, context)
+
+ if api_key:
+ os.environ['MINIMAX_API_KEY'] = api_key
+
+ def get_instructions_for_objective(self, original_user_request: str, step_num: int = 0) -> dict[str, Any]:
+ message: list[dict[str, Any]] = self.format_user_request_for_llm(original_user_request, step_num)
+ llm_response = self.send_message_to_llm(message)
+ json_instructions: dict[str, Any] = self.convert_llm_response_to_json_instructions(llm_response)
+ return json_instructions
+
+ def format_user_request_for_llm(self, original_user_request, step_num) -> list[dict[str, Any]]:
+ base64_img: str = Screen().get_screenshot_in_base64()
+
+ request_data: str = json.dumps({
+ 'original_user_request': original_user_request,
+ 'step_num': step_num
+ })
+
+ message = [
+ {'type': 'text', 'text': self.context + request_data},
+ {'type': 'image_url',
+ 'image_url': {
+ 'url': f'data:image/jpeg;base64,{base64_img}'
+ }
+ }
+ ]
+
+ return message
+
+ def send_message_to_llm(self, message):
+ response = self.client.chat.completions.create(
+ model=self.model_name,
+ messages=[
+ {
+ 'role': 'user',
+ 'content': message,
+ }
+ ],
+ max_tokens=800,
+ temperature=min(max(0.01, 0.7), 1.0),
+ )
+ return response
+
+ def convert_llm_response_to_json_instructions(self, llm_response) -> dict[str, Any]:
+ llm_response_data: str = llm_response.choices[0].message.content.strip()
+
+ # Strip MiniMax thinking tags if present (M2.5+ models may include <think>...</think> blocks)
+ if '<think>' in llm_response_data and '</think>' in llm_response_data:
+ think_end = llm_response_data.rfind('</think>')
+ llm_response_data = llm_response_data[think_end + len('</think>'):].strip()
+
+ start_index = llm_response_data.find('{')
+ end_index = llm_response_data.rfind('}')
+
+ try:
+ json_response = json.loads(llm_response_data[start_index:end_index + 1].strip())
+ except Exception as e:
+ print(f'Error while parsing JSON response - {e}')
+ json_response = {}
+
+ return json_response

app/ui.py

@@ -69,6 +69,11 @@ def create_widgets(self) -> None:
  ('Gemini gemini-3-flash-preview', 'gemini-3-flash-preview'),
  ]
 
+ minimax_models = [
+ ('MiniMax-M2.7 (1M Context)', 'MiniMax-M2.7'),
+ ('MiniMax-M2.5-highspeed (204K, Fast)', 'MiniMax-M2.5-highspeed'),
+ ]
+
  deprecated_models = [
  ('GPT-4o (Medium-Accurate, Medium-Fast)', 'gpt-4o'),
  ('GPT-4o-mini (Cheapest, Fastest)', 'gpt-4o-mini'),
@@ -93,6 +98,12 @@ def create_widgets(self) -> None:
  ttk.Radiobutton(radio_frame, text=text, value=value, variable=self.model_var, bootstyle="info").pack(
  anchor=ttk.W, pady=5)
 
+ ttk.Separator(radio_frame, orient='horizontal').pack(fill='x', pady=8)
+
+ for text, value in minimax_models:
+ ttk.Radiobutton(radio_frame, text=text, value=value, variable=self.model_var, bootstyle="info").pack(
+ anchor=ttk.W, pady=5)
+
  ttk.Separator(radio_frame, orient='horizontal').pack(fill='x', pady=10)
 
  self.deprecated_expanded = False
@@ -192,7 +203,7 @@ def __init__(self, parent):
 
  def create_widgets(self) -> None:
  # API Key Widgets
- label_api = ttk.Label(self, text='OpenAI/Gemini/LLM Model API Key:', bootstyle="info")
+ label_api = ttk.Label(self, text='OpenAI/Gemini/MiniMax/LLM Model API Key:', bootstyle="info")
  label_api.pack(pady=10)
  self.api_key_entry = ttk.Entry(self, width=30)
  self.api_key_entry.pack()

README.md

@@ -7,7 +7,7 @@
 ### Control Your Computer Using LLMs
 
 Open Interface
-- Self-drives your computer by sending your requests to an LLM backend (GPT-4o, Gemini, etc) to figure out the required steps.
+- Self-drives your computer by sending your requests to an LLM backend (GPT-4o, Gemini, MiniMax, etc) to figure out the required steps.
 - Automatically executes these steps by simulating keyboard and mouse input.
 - Course-corrects by sending the LLM backend updated screenshots of the progress as needed.
 
@@ -172,6 +172,17 @@ Open Interface
 
 </details>
 
+<details>
+ <summary><b>Set up the MiniMax API key</b></summary>
+
+- Go to Settings -> Advanced Settings and select the MiniMax model you wish to use (MiniMax-M2.7 or MiniMax-M2.5-highspeed).
+- Get your MiniMax API key from https://platform.minimaxi.com/
+- Save the API key in Open Interface settings.
+- Save the settings and **restart the app**.
+- Alternatively, set the `MINIMAX_API_KEY` environment variable and the API key will be auto-detected.
+
+</details>
+
 <details>
  <summary><b>Optional: Setup a Custom LLM</b></summary>
 

app/models/factory.py

@@ -1,6 +1,7 @@
 from models.gpt4o import GPT4o
 from models.gpt4v import GPT4v
 from models.gpt5 import GPT5
+from models.minimax import MiniMax
 from models.openai_computer_use import OpenAIComputerUse
 from models.gemini import Gemini
 
@@ -19,6 +20,8 @@ def create_model(model_name, *args):
  return GPT4v(model_name, *args)
  elif model_name.startswith("gemini"):
  return Gemini(model_name, *args[1:])
+ elif model_name.startswith("MiniMax"):
+ return MiniMax(model_name, *args)
  else:
  # Llama/Llava models will work with the standard code I wrote for GPT4V without the assitant mode features of gpt4o
  return GPT4v(model_name, *args)

app/models/minimax.py

@@ -0,0 +1,87 @@
+import json
+import os
+from typing import Any
+
+from models.model import Model
+from utils.screen import Screen
+
+MINIMAX_BASE_URL = 'https://api.minimax.io/v1/'
+
+
+class MiniMax(Model):
+ """
+ MiniMax model integration via OpenAI-compatible API.
+ Supports MiniMax-M2.7 (latest) and MiniMax-M2.5-highspeed (204K context).
+ """
+
+ def __init__(self, model_name, base_url, api_key, context):
+ # Use MiniMax base URL unless user has explicitly set a custom one
+ if not base_url or base_url.rstrip('/') + '/' == 'https://api.openai.com/v1/':
+ base_url = MINIMAX_BASE_URL
+
+ # Prefer MINIMAX_API_KEY env var if no API key was provided
+ if not api_key:
+ api_key = os.environ.get('MINIMAX_API_KEY', '')
+
+ super().__init__(model_name, base_url, api_key, context)
+
+ if api_key:
+ os.environ['MINIMAX_API_KEY'] = api_key
+
+ def get_instructions_for_objective(self, original_user_request: str, step_num: int = 0) -> dict[str, Any]:
+ message: list[dict[str, Any]] = self.format_user_request_for_llm(original_user_request, step_num)
+ llm_response = self.send_message_to_llm(message)
+ json_instructions: dict[str, Any] = self.convert_llm_response_to_json_instructions(llm_response)
+ return json_instructions
+
+ def format_user_request_for_llm(self, original_user_request, step_num) -> list[dict[str, Any]]:
+ base64_img: str = Screen().get_screenshot_in_base64()
+
+ request_data: str = json.dumps({
+ 'original_user_request': original_user_request,
+ 'step_num': step_num
+ })
+
+ message = [
+ {'type': 'text', 'text': self.context + request_data},
+ {'type': 'image_url',
+ 'image_url': {
+ 'url': f'data:image/jpeg;base64,{base64_img}'
+ }
+ }
+ ]
+
+ return message
+
+ def send_message_to_llm(self, message):
+ response = self.client.chat.completions.create(
+ model=self.model_name,
+ messages=[
+ {
+ 'role': 'user',
+ 'content': message,
+ }
+ ],
+ max_tokens=800,
+ temperature=min(max(0.01, 0.7), 1.0),
+ )
+ return response
+
+ def convert_llm_response_to_json_instructions(self, llm_response) -> dict[str, Any]:
+ llm_response_data: str = llm_response.choices[0].message.content.strip()
+
+ # Strip MiniMax thinking tags if present (M2.5+ models may include <think>...</think> blocks)
+ if '<think>' in llm_response_data and '</think>' in llm_response_data:
+ think_end = llm_response_data.rfind('</think>')
+ llm_response_data = llm_response_data[think_end + len('</think>'):].strip()
+
+ start_index = llm_response_data.find('{')
+ end_index = llm_response_data.rfind('}')
+
+ try:
+ json_response = json.loads(llm_response_data[start_index:end_index + 1].strip())
+ except Exception as e:
+ print(f'Error while parsing JSON response - {e}')
+ json_response = {}
+
+ return json_response

app/ui.py

@@ -69,6 +69,11 @@ def create_widgets(self) -> None:
  ('Gemini gemini-3-flash-preview', 'gemini-3-flash-preview'),
  ]
 
+ minimax_models = [
+ ('MiniMax-M2.7 (1M Context)', 'MiniMax-M2.7'),
+ ('MiniMax-M2.5-highspeed (204K, Fast)', 'MiniMax-M2.5-highspeed'),
+ ]
+
  deprecated_models = [
  ('GPT-4o (Medium-Accurate, Medium-Fast)', 'gpt-4o'),
  ('GPT-4o-mini (Cheapest, Fastest)', 'gpt-4o-mini'),
@@ -93,6 +98,12 @@ def create_widgets(self) -> None:
  ttk.Radiobutton(radio_frame, text=text, value=value, variable=self.model_var, bootstyle="info").pack(
  anchor=ttk.W, pady=5)
 
+ ttk.Separator(radio_frame, orient='horizontal').pack(fill='x', pady=8)
+
+ for text, value in minimax_models:
+ ttk.Radiobutton(radio_frame, text=text, value=value, variable=self.model_var, bootstyle="info").pack(
+ anchor=ttk.W, pady=5)
+
  ttk.Separator(radio_frame, orient='horizontal').pack(fill='x', pady=10)
 
  self.deprecated_expanded = False
@@ -192,7 +203,7 @@ def __init__(self, parent):
 
  def create_widgets(self) -> None:
  # API Key Widgets
- label_api = ttk.Label(self, text='OpenAI/Gemini/LLM Model API Key:', bootstyle="info")
+ label_api = ttk.Label(self, text='OpenAI/Gemini/MiniMax/LLM Model API Key:', bootstyle="info")
  label_api.pack(pady=10)
  self.api_key_entry = ttk.Entry(self, width=30)
  self.api_key_entry.pack()