ismail/haikal
1
0
Fork 0
Code Issues 2 Pull Requests Actions Packages Projects Releases Wiki Activity

Merge remote-tracking branch 'origin/main'

Browse Source
This commit is contained in:
Marwan Alwali 2025-05-26 16:37:30 +03:00
commit 97ebf23732
2 changed files with 15 additions and 13 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

26
haikalbot/Optimizing Qwen3-8B for Arabic Language Support in Django AI Analyst.md
View File

@ -18,7 +18,7 @@ Update your Django settings to use Qwen3-8B:
```python ```python
# In settings.py # In settings.py
OLLAMA_BASE_URL = "http://localhost:11434" OLLAMA_BASE_URL = "http://10.10.1.132:11434"
OLLAMA_MODEL = "qwen3:8b" OLLAMA_MODEL = "qwen3:8b"
OLLAMA_TIMEOUT = 120 # Seconds OLLAMA_TIMEOUT = 120 # Seconds
``` ```
@ -34,10 +34,10 @@ def get_ollama_llm():
""" """
try: try:
# Get settings from Django settings or use defaults # Get settings from Django settings or use defaults
base_url = getattr(settings, 'OLLAMA_BASE_URL', 'http://localhost:11434') base_url = getattr(settings, 'OLLAMA_BASE_URL', 'http://10.10.1.132:11434')
model = getattr(settings, 'OLLAMA_MODEL', 'qwen3:8b') model = getattr(settings, 'OLLAMA_MODEL', 'qwen3:8b')
timeout = getattr(settings, 'OLLAMA_TIMEOUT', 120) timeout = getattr(settings, 'OLLAMA_TIMEOUT', 120)
# Configure Ollama with parameters optimized for Qwen3-8B with Arabic # Configure Ollama with parameters optimized for Qwen3-8B with Arabic
return Ollama( return Ollama(
base_url=base_url, base_url=base_url,
@ -71,7 +71,7 @@ def create_prompt_analyzer_chain(language='ar'):
llm = get_ollama_llm() llm = get_ollama_llm()
if not llm: if not llm:
return None return None
# Define the prompt template optimized for Qwen3-8B # Define the prompt template optimized for Qwen3-8B
if language == 'ar': if language == 'ar':
template = """ template = """
@ -79,9 +79,9 @@ def create_prompt_analyzer_chain(language='ar'):
1. نوع التحليل المطلوب 1. نوع التحليل المطلوب
2. نماذج البيانات المستهدفة 2. نماذج البيانات المستهدفة
3. أي معلمات استعلام 3. أي معلمات استعلام
الاستعلام: {prompt} الاستعلام: {prompt}
قم بتقديم إجابتك بتنسيق JSON فقط، بدون أي نص إضافي، كما يلي: قم بتقديم إجابتك بتنسيق JSON فقط، بدون أي نص إضافي، كما يلي:
```json ```json
{{ {{
@ -97,9 +97,9 @@ def create_prompt_analyzer_chain(language='ar'):
1. The type of analysis required 1. The type of analysis required
2. Target data models 2. Target data models
3. Any query parameters 3. Any query parameters
Prompt: {prompt} Prompt: {prompt}
Provide your answer in JSON format only, without any additional text, as follows: Provide your answer in JSON format only, without any additional text, as follows:
```json ```json
{ {
@ -109,13 +109,13 @@ def create_prompt_analyzer_chain(language='ar'):
} }
``` ```
""" """
# Create the prompt template # Create the prompt template
prompt_template = PromptTemplate( prompt_template = PromptTemplate(
input_variables=["prompt"], input_variables=["prompt"],
template=template template=template
) )
# Create and return the LLM chain # Create and return the LLM chain
return LLMChain(llm=llm, prompt=prompt_template) return LLMChain(llm=llm, prompt=prompt_template)
``` ```
@ -135,13 +135,13 @@ def _parse_llm_json_response(result):
if json_match: if json_match:
json_str = json_match.group(1).strip() json_str = json_match.group(1).strip()
return json.loads(json_str) return json.loads(json_str)
# If no markdown blocks, try to find JSON object directly # If no markdown blocks, try to find JSON object directly
json_match = re.search(r'({[\s\S]*})', result) json_match = re.search(r'({[\s\S]*})', result)
if json_match: if json_match:
json_str = json_match.group(1).strip() json_str = json_match.group(1).strip()
return json.loads(json_str) return json.loads(json_str)
# If still no match, try to parse the entire response as JSON # If still no match, try to parse the entire response as JSON
return json.loads(result.strip()) return json.loads(result.strip())
except Exception as e: except Exception as e:
@ -159,7 +159,7 @@ def _parse_llm_json_response(result):
## Handling Arabic-Specific Challenges with Qwen3-8B ## Handling Arabic-Specific Challenges with Qwen3-8B
1. **Diacritics**: Qwen3-8B handles Arabic diacritics well, but for consistency, consider normalizing input by removing diacritics 1. **Diacritics**: Qwen3-8B handles Arabic diacritics well, but for consistency, consider normalizing input by removing diacritics
2. **Text Direction**: When displaying results in frontend, ensure proper RTL (right-to-left) support 2. **Text Direction**: When displaying results in frontend, ensure proper RTL (right-to-left) support
3. **Dialectal Variations**: Qwen3-8B performs best with Modern Standard Arabic (MSA), but has reasonable support for major dialects 3. **Dialectal Variations**: Qwen3-8B performs best with Modern Standard Arabic (MSA), but has reasonable support for major dialects

2
requirements_dev.txt
View File

@ -92,3 +92,5 @@ typing_extensions==4.13.0
tzdata==2025.2 tzdata==2025.2
urllib3==2.3.0 urllib3==2.3.0
wcwidth==0.2.13 wcwidth==0.2.13
langchain
langchain_ollama