DeepSeek AI Model

[srikanthsp20]

Select Topic Area

General

Body

The deepseek ai model is showing 'Unknown model' error ijn Pydantic examples.....Openai & gemini are working fine..can someone reply the solution......thks
[email protected]
21-March-2025

[shinybrightstar]

Welcome to the GitHub Community, @srikanthsp20 , we're happy you're here! You are more likely to get a useful response if you are posting your question(s) in the applicable category and are explicit about what your project entails--giving a few more details might help someone give you a nudge in the right direction. I've gone ahead and moved it for you. Good luck!

[srikanthsp20]

The OpenAI / Gemini Model or engine is working fine but indicates "Insufficient Quota"..I want to use deepseek which is free, in Pydantic examples but 'Unknown Model' error is displayed...can u reply the solution..thks

[lehoanganh75]

oh thanks

[srikanthsp20]

The OpenAI / Gemini Model or engine is working fine but indicates "Insufficient Quota"..I want to use deepseek which is free, in Pydantic examples but 'Unknown Model' error is displayed...can u reply the solution..thks

[Edmon02]

DeepSeek AI is a cutting-edge model focusing on generative AI and large language models (LLMs). If you are trying to integrate or fine-tune it, here are a few things to check:
• Installation: Ensure you have the correct dependencies using pip install deepseek.
• Model Access: Some models require API keys. Verify if you need one from the official DeepSeek AI website.
• Fine-Tuning: If you’re looking to fine-tune, use libraries like Hugging Face Transformers or PyTorch.

Example usage:

from transformers import AutoModelForCausalLM, AutoTokenizer
 
model_name = "deepseek/deepseek-ai"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)
 
inputs = tokenizer("Hello, how can I help you?", return_tensors="pt")
outputs = model.generate(**inputs)
print(tokenizer.decode(outputs[0]))

Let me know what specific issue you’re facing, and I’ll be happy to help!

[srikanthsp20]

hello Edmon02,
The above code is for uploading the image through Android's camera...what i want to do is select an area on the screen and get the audio speech output in the android.. do u think its possible? if so, can u share the code,..thks

[srikanthsp20]

***** select an area on the screen** Using finger on the touchphone screen---

[Edmon02]

hello Edmon02, The above code is for uploading the image through Android's camera...what i want to do is select an area on the screen and get the audio speech output in the android.. do u think its possible? if so, can u share the code,..thks

Hi! Thanks for clarifying your requirement—I understand now that you want to select a specific area on the screen (e.g., a portion of an image or text displayed on your Android phone) using your finger on the touchscreen, extract the text from that area, and convert it to audio speech output. Yes, this is definitely possible! Below, I’ll explain how to approach it and provide a sample code snippet using Python and Kivy, since Kivy is well-suited for handling touch events on Android.

How It Can Work:

1. Display the Image: Show an image (e.g., a train timetable) on the screen.
2. Touch Selection: Allow the user to drag their finger to draw a rectangle over the desired area.
3. Crop the Selected Area: Extract that portion of the image.
4. OCR: Use Tesseract to recognize text in the cropped area.
5. Text-to-Speech: Convert the extracted text to audio using gTTS or another TTS library.
6. Android Deployment: Package it as an APK.

Challenges:

• Kivy handles touch events well, but you’ll need to map the screen coordinates to the image’s coordinates.
• The DeepSeek model might not be necessary here unless you want to refine the extracted text further (I’ll keep it optional).

Sample Code:

Here’s a basic Kivy app that lets you select an area on an image with your finger and converts the extracted text to speech:

from kivy.app import App
from kivy.uix.widget import Widget
from kivy.uix.image import Image
from kivy.graphics import Line, Rectangle
from kivy.core.window import Window
from kivy.uix.button import Button
from kivy.uix.boxlayout import BoxLayout
import pytesseract
from PIL import Image as PILImage
from gtts import gTTS
import os

class TouchWidget(Widget):
    def __init__(self, **kwargs):
        super(TouchWidget, self).__init__(**kwargs)
        self.start_pos = None
        self.end_pos = None
        self.image_path = "timetable.jpg"  # Replace with your image path
        self.img = PILImage.open(self.image_path)

        # Add the image to the widget
        with self.canvas:
            self.image = Image(source=self.image_path, size=Window.size)

    def on_touch_down(self, touch):
        # Record the starting position of the touch
        self.start_pos = (touch.x, touch.y)
        with self.canvas:
            self.rect = Rectangle(pos=self.start_pos, size=(1, 1))
    
    def on_touch_move(self, touch):
        # Update the rectangle as the finger moves
        self.end_pos = (touch.x, touch.y)
        with self.canvas:
            self.canvas.remove(self.rect)
            width = self.end_pos[0] - self.start_pos[0]
            height = self.end_pos[1] - self.start_pos[1]
            self.rect = Rectangle(pos=self.start_pos, size=(width, height))

    def on_touch_up(self, touch):
        # When the finger is lifted, process the selected area
        self.end_pos = (touch.x, touch.y)
        self.process_selection()

    def process_selection(self):
        # Crop the selected area from the image
        x1, y1 = self.start_pos
        x2, y2 = self.end_pos
        # Convert Kivy coordinates (bottom-left origin) to PIL (top-left origin)
        img_height = self.img.height
        crop_box = (min(x1, x2), img_height - max(y1, y2), max(x1, x2), img_height - min(y1, y2))
        cropped_img = self.img.crop(crop_box)

        # Extract text using OCR
        text = pytesseract.image_to_string(cropped_img)
        print(f"Extracted Text: {text}")

        # Convert text to speech
        if text.strip():
            tts = gTTS(text=text, lang="en")  # Change "en" to "ta" for Tamil, etc.
            tts.save("output.mp3")
            # Play the audio (you may need a media player library like `kivy.core.audio`)
            from kivy.core.audio import SoundLoader
            sound = SoundLoader.load("output.mp3")
            if sound:
                sound.play()
        else:
            print("No text detected in the selected area.")

class MyApp(App):
    def build(self):
        layout = BoxLayout(orientation="vertical")
        touch_widget = TouchWidget()
        layout.add_widget(touch_widget)
        return layout

if __name__ == "__main__":
    MyApp().run()

How to Use This Code:

1. Install Dependencies:
   • On your PC: pip install kivy pytesseract pillow gtts
   • Install Tesseract OCR on your system (e.g., sudo apt install tesseract-ocr on Linux).
2. Prepare an Image:
   • Replace "timetable.jpg" with the path to your image file (e.g., a train timetable photo).
3. Run on PC First:
   • Test the code on your PC. Touch and drag with your mouse to select an area, and it’ll extract text and play it as speech.
4. Deploy to Android:
   • Use Buildozer (pip install buildozer) to package it into an APK:
     • Create a buildozer.spec file with buildozer init.
     • Edit the spec file to include dependencies: requirements = kivy, pillow, pytesseract, gtts.
     • Run buildozer android debug to generate the APK.
   • Install the APK on your Android phone and use your finger to select the area.

Notes:

• Image Source: This code assumes a preloaded image (timetable.jpg). To use the camera, you’d need to integrate Kivy’s Camera widget or Chaquopy with Android’s camera API—let me know if you want that instead!
• Multilingual Support: Change the lang parameter in gTTS (e.g., "ta" for Tamil, "hi" for Hindi).
• DeepSeek Optional: If you want to use DeepSeek to refine the OCR text, add it before the TTS step (like in my previous reply), but it might need an API for mobile use.

Let me know if you run into issues setting this up or need help with camera integration instead of a static image! I’m happy to refine the code further.

[srikanthsp20]

Thks very much for your time.....

self.image_path = "timetable.jpg" # Replace with your image path

Again the Image is in a .jpg file format, What I wanted was to select text area from any image (From URL/Https:/Websites- browser images)
and convert the text area to speech.. do u think its possible? Thks
SPSrikanth Chennai India
4-4-2025

[srikanthsp20]

Saw Chat AI -it gives an example code of 'Frame Grab after time delay & convert to .jpg....but i want an Real-time frame grab and converting the text area to speech... dou think it can be done?...Thks for ur time

[srikanthsp20]

Hello Edmon02 I ran the python program offline on my Notebook i3 64bit Intel(R) HD graphics 3000 and it showed "Fatal kivy Error"-----"OpenGL 2.0 Not found" ......I updated the graphics driver from Microsoft site but still same error...... Kindly reply if its a Hardware problem or some other solution is possible....thks for sparing your time.. SPSrikanth Chennai India

…

On Thu, Apr 3, 2025 at 10:42 PM Edmon Sahakyan ***@***.***> wrote: hello Edmon02, The above code is for uploading the image through Android's camera...what i want to do is select an area on the screen and get the audio speech output in the android.. do u think its possible? if so, can u share the code,..thks Hi! Thanks for clarifying your requirement—I understand now that you want to select a specific area on the screen (e.g., a portion of an image or text displayed on your Android phone) using your finger on the touchscreen, extract the text from that area, and convert it to audio speech output. Yes, this is definitely possible! Below, I’ll explain how to approach it and provide a sample code snippet using Python and Kivy, since Kivy is well-suited for handling touch events on Android. How It Can Work: 1. *Display the Image:* Show an image (e.g., a train timetable) on the screen. 2. *Touch Selection:* Allow the user to drag their finger to draw a rectangle over the desired area. 3. *Crop the Selected Area:* Extract that portion of the image. 4. *OCR:* Use Tesseract to recognize text in the cropped area. 5. *Text-to-Speech:* Convert the extracted text to audio using gTTS or another TTS library. 6. *Android Deployment:* Package it as an APK. Challenges: - Kivy handles touch events well, but you’ll need to map the screen coordinates to the image’s coordinates. - The DeepSeek model might not be necessary here unless you want to refine the extracted text further (I’ll keep it optional). Sample Code: Here’s a basic Kivy app that lets you select an area on an image with your finger and converts the extracted text to speech: from kivy.app import Appfrom kivy.uix.widget import Widgetfrom kivy.uix.image import Imagefrom kivy.graphics import Line, Rectanglefrom kivy.core.window import Windowfrom kivy.uix.button import Buttonfrom kivy.uix.boxlayout import BoxLayoutimport pytesseractfrom PIL import Image as PILImagefrom gtts import gTTSimport os class TouchWidget(Widget): def __init__(self, **kwargs): super(TouchWidget, self).__init__(**kwargs) self.start_pos = None self.end_pos = None self.image_path = "timetable.jpg" # Replace with your image path self.img = PILImage.open(self.image_path) # Add the image to the widget with self.canvas: self.image = Image(source=self.image_path, size=Window.size) def on_touch_down(self, touch): # Record the starting position of the touch self.start_pos = (touch.x, touch.y) with self.canvas: self.rect = Rectangle(pos=self.start_pos, size=(1, 1)) def on_touch_move(self, touch): # Update the rectangle as the finger moves self.end_pos = (touch.x, touch.y) with self.canvas: self.canvas.remove(self.rect) width = self.end_pos[0] - self.start_pos[0] height = self.end_pos[1] - self.start_pos[1] self.rect = Rectangle(pos=self.start_pos, size=(width, height)) def on_touch_up(self, touch): # When the finger is lifted, process the selected area self.end_pos = (touch.x, touch.y) self.process_selection() def process_selection(self): # Crop the selected area from the image x1, y1 = self.start_pos x2, y2 = self.end_pos # Convert Kivy coordinates (bottom-left origin) to PIL (top-left origin) img_height = self.img.height crop_box = (min(x1, x2), img_height - max(y1, y2), max(x1, x2), img_height - min(y1, y2)) cropped_img = self.img.crop(crop_box) # Extract text using OCR text = pytesseract.image_to_string(cropped_img) print(f"Extracted Text: {text}") # Convert text to speech if text.strip(): tts = gTTS(text=text, lang="en") # Change "en" to "ta" for Tamil, etc. tts.save("output.mp3") # Play the audio (you may need a media player library like `kivy.core.audio`) from kivy.core.audio import SoundLoader sound = SoundLoader.load("output.mp3") if sound: sound.play() else: print("No text detected in the selected area.") class MyApp(App): def build(self): layout = BoxLayout(orientation="vertical") touch_widget = TouchWidget() layout.add_widget(touch_widget) return layout if __name__ == "__main__": MyApp().run() How to Use This Code: 1. *Install Dependencies:* - On your PC: pip install kivy pytesseract pillow gtts - Install Tesseract OCR on your system (e.g., sudo apt install tesseract-ocr on Linux). 2. *Prepare an Image:* - Replace "timetable.jpg" with the path to your image file (e.g., a train timetable photo). 3. *Run on PC First:* - Test the code on your PC. Touch and drag with your mouse to select an area, and it’ll extract text and play it as speech. 4. *Deploy to Android:* - Use *Buildozer* (pip install buildozer) to package it into an APK: - Create a buildozer.spec file with buildozer init. - Edit the spec file to include dependencies: requirements = kivy, pillow, pytesseract, gtts. - Run buildozer android debug to generate the APK. - Install the APK on your Android phone and use your finger to select the area. Notes: - *Image Source:* This code assumes a preloaded image (timetable.jpg). To use the camera, you’d need to integrate Kivy’s Camera widget or Chaquopy with Android’s camera API—let me know if you want that instead! - *Multilingual Support:* Change the lang parameter in gTTS (e.g., "ta" for Tamil, "hi" for Hindi). - *DeepSeek Optional:* If you want to use DeepSeek to refine the OCR text, add it before the TTS step (like in my previous reply), but it might need an API for mobile use. Let me know if you run into issues setting this up or need help with camera integration instead of a static image! I’m happy to refine the code further. — Reply to this email directly, view it on GitHub <#154582 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/BJ2YREQ4GA6EGBJZJFCDVLD2XVTX7AVCNFSM6AAAAABZPTYGWOVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTENZRGY4TOOI> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[Edmon02]

Hi! Thanks for the update, and no worries at all—I’m happy to help you troubleshoot this. The "Fatal Kivy Error: OpenGL 2.0 Not Found" issue you’re seeing is a common one when running Kivy on a system, and it’s tied to graphics support. Let’s break it down and find a solution.

What’s Causing the Error?

Kivy relies on OpenGL (version 2.0 or higher) for rendering its graphical interface. The error means your system either:

1. Doesn’t support OpenGL 2.0 due to hardware limitations, or
2. Has the capability but lacks the right software/drivers to enable it.

You mentioned your setup:

• Notebook i3 64-bit (likely an older Intel Core i3 processor).
• Intel HD Graphics 3000 (integrated GPU, typically found in 2nd-gen Intel Core processors from around 2011).
• Updated graphics driver from Microsoft’s site, but the error persists.

Is It a Hardware Problem?

The Intel HD Graphics 3000 does support OpenGL 2.0 (and even up to OpenGL 3.1 on Windows), so this shouldn’t be a hardware limitation. However:

• Driver Issue: Microsoft’s default drivers (via Windows Update) are often basic and may not fully enable OpenGL support.
• Old Hardware: While the GPU supports OpenGL 2.0, compatibility can depend on the operating system and driver version.

Solutions to Try:

Here are a few steps to resolve this, starting with the most likely fixes:

1. Install Intel’s Official Driver (Not Microsoft’s):

   • Microsoft’s drivers are generic and often lack full OpenGL support. Download the driver directly from Intel:
     • Go to Intel’s Download Center: https://www.intel.com/content/www/us/en/support/detect.html
     • Use the "Intel Driver & Support Assistant" to auto-detect your hardware, or manually select "Intel HD Graphics 3000" and your OS (e.g., Windows 10 64-bit).
     • Install the driver and restart your notebook.
   • After updating, test the Kivy code again.

2. Check OpenGL Version:

   • To confirm what your system supports, download a tool like OpenGL Extensions Viewer (free from realtech-vr.com) and run it. It’ll show your OpenGL version. If it’s below 2.0 even after the driver update, we’ll need a workaround.

3. Force Kivy to Use a Software Renderer:

   • If the driver fix doesn’t work, you can tell Kivy to bypass hardware OpenGL and use a software-based renderer (slower but should work):
     • Before running the app, set this environment variable in your Python script:

       import os
       os.environ["KIVY_GL_BACKEND"] = "angle_sdl2"  # Use ANGLE renderer
       # OR
       os.environ["KIVY_GL_BACKEND"] = "sdl2"  # Default SDL2, may work better

     • Add this at the top of the code I shared, then run it again.
   • Alternatively, run the script from the command line with:

     set KIVY_GL_BACKEND=angle_sdl2 && python your_script.py
     

4. Update Your Operating System:

   • If you’re on an older OS (e.g., Windows 7), OpenGL support might be limited. Upgrading to Windows 10 (if not already) could help, as Intel provides better driver support for newer OS versions.

5. Fallback Option: Use a Different Machine or Emulator:

   • If the above don’t work and it’s a persistent hardware/driver issue, test the code on:
     • Another PC with a newer GPU.
     • An Android emulator (e.g., Android Studio’s AVD) on your notebook, though this requires building the APK with Buildozer first.

Next Steps:

• Try Step 1 (Intel driver) first—it’s the most likely fix since your HD Graphics 3000 should support OpenGL 2.0.
• Let me know your OS (e.g., Windows 10, 7) and the result after installing Intel’s driver. If it still fails, I can tweak the code further or suggest an alternative framework that doesn’t need OpenGL (though Kivy is ideal for Android touch support).

This is likely a software/driver issue rather than a hardware limitation, so we should be able to get it running! Let me know how it goes.

[srikanthsp20]

Hello,
Good Day Again- The same error- "Found OpenGL 1.1" and "Not found OpenGL 2.0" is displaying
"Upgrade Hardware and/or drivers" message
I dont have another system to test...will an USB adapter to secondary Monitor/TV work as a substitute for OpenGL 2.0??
Thks for your time
Srikanth SP
Chennai India

[srikanthsp20]

Hello
My oversight- Its already available on Android-Display text to speech (Multilingual) in the 'Accessibility' settings
Any way thks
SPSrikanth
Chennai India

[ji603]

Hello
My oversight- Its already available on Android-Display text to speech (Multilingual) in the 'Accessibility' settings
Any way thks
SPSrikanth
Chennai India

Congrats 🎉

[ShubhamV2503]

The "Unknown model" error in DeepSeek-AI with Pydantic likely means the model name isn't supported or integrated properly. Ensure you're using a valid model name and that the DeepSeek API is correctly configured in your client.

[nikhilkr16]

It looks like you're getting a "Unknown model" error when using DeepSeek AI with Pydantic examples, while OpenAI and Gemini work fine.

✅ Possible Fixes:

Check model name spelling: DeepSeek might require a specific model name (e.g., "deepseek-coder" or "deepseek-chat"). Make sure you're using the exact model ID they support.

Update your SDK: If you're using LangChain, LlamaIndex, or any wrapper — update the package. Support for DeepSeek is new and may not be fully integrated in older versions.

API compatibility: DeepSeek’s API might not fully support structured outputs like Pydantic yet. Try testing with a basic prompt first to confirm it’s working.

Fallback workaround: Instead of using PydanticOutputParser, try parsing the raw JSON manually as a temporary solution.

🔧 Example fix:

If the raw output is not valid JSON, wrap your prompt like:
"Give the response in this exact JSON format: {"name": "John", "age": 25}"

[yuyuayush]

try to find more Reasearch analysis

[zenriquezs]

Hi Srikanth,

Thanks for sharing your detailed updates — I admire your persistence!

Regarding your latest question:

“Will a USB adapter to a secondary Monitor/TV work as a substitute for OpenGL 2.0?”

Unfortunately, a USB-to-HDMI or USB display adapter won’t solve the OpenGL 2.0 limitation, because the adapter doesn’t improve the GPU capabilities or driver support of your internal graphics card (Intel HD Graphics 3000). OpenGL is handled by the GPU and its driver, not by the external display output.

Here are a few options to consider:

• Try software rendering in Kivy (fallback mode):
  This often works even when OpenGL 2.0 isn’t available:

python
Copiar
Editar
import os
os.environ["KIVY_GL_BACKEND"] = "angle_sdl2" # or "sdl2"

• Use Android’s built-in accessibility (as you discovered):
  If your main goal is to extract and read text aloud from the screen, Android's Text-to-Speech in Accessibility Settings is a great low-hardware solution.

• Alternative idea:
  If your goal is real-time screen capture + OCR + TTS on Android, consider using:

ML Kit for on-device OCR

Android’s built-in TTS

Developed via Java/Kotlin or via Python+Chaquopy (if you still prefer Python)

Let me know if you need help setting up a lightweight mobile app or exploring OCR alternatives that don’t depend on OpenGL.

[ctcrahul]

i Will agree with you

[Sukhtumur]

“DeepSeek models aren’t yet officially integrated into the OpenAI or Pydantic schemas. That ‘Unknown model’ error means the client library doesn’t recognize the model ID. You need to either:

1. Use an SDK that supports DeepSeek explicitly,
2. Or register the model manually in your client setup before calling it.

If you’re using the OpenAI Python client, ensure the base URL points to DeepSeek’s endpoint and not api.openai.com. Otherwise, Pydantic validation will fail since ‘deepseek’ isn’t in the known model list.”

[nitin-999-code]

The “Unknown model” error in Pydantic AI appears because the library doesn’t recognize the model name you gave for DeepSeek. This usually happens if:

You’re using the wrong model name (like "deepseek-chat" instead of the exact one, e.g. "deepseek-chat-v2").

The DeepSeek provider isn’t set correctly.

Your Pydantic AI version is outdated.

Fix:

Update Pydantic AI

pip install --upgrade pydantic_ai

Use the correct model and provider

from pydantic_ai.models.openai import OpenAIChatModel
from pydantic_ai.providers.deepseek import DeepSeekProvider

model = OpenAIChatModel(
"deepseek-chat-v2", # use exact supported model name
provider=DeepSeekProvider(api_key="YOUR_DEEPSEEK_API_KEY")
)

Ensure API key and endpoint are valid (check your DeepSeek dashboard).

[sanatan0511]

I think it is due to old age software and hardware it is do not supoted .or old version of software .for ex some services of youtube not run in android 7 or 8 so need to upgrade same with 10 also.or might be low version of software please upgrade it :)

[dimpal-yadav]

Select Topic Area

General

Body

The deepseek ai model is showing 'Unknown model' error ijn Pydantic examples.....Openai & gemini are working fine..can someone reply the solution......thks

Make sure you’re using the correct DeepSeek model ID and have configured the DeepSeek provider in Pydantic AI. Updating to the latest Pydantic AI version often fixes the “Unknown model” error.