fix: merge system messages to comply with Qwen vLLM chat template

[OlegSob-glitch]

Summary:

• Fixes 'System message must be at the beginning' error when using Qwen models via vLLM
• Qwen's chat template requires exactly ONE system message at position 0 of messages array

Changes:

1. deeptutor/agents/chat/agentic_pipeline.py: Merge system_prompt and memory_context into single system message, filter out system messages from conversation_history
2. deeptutor/services/session/context_builder.py: Filter out system messages from DB since summary is already added as system
3. deeptutor/agents/chat/chat_agent.py: Merge system prompt and RAG context into single system message

Benefits:

• Fixes Qwen vLLM compatibility issue completely
• Reduces token usage by eliminating duplicate system messages
• Does not break any functionality - system context from DB is already in compressed_summary
• Cleaner message structure with exactly one system message at position 0

Testing:

• Verified with Qwen model via vLLM - no more template errors
• Chat with memory files (PROFILE.md/SUMMARY.md) works correctly
• Chat with conversation history works correctly
• No regression with other models (GPT-4, etc.)

Description

A clear and concise description of the changes.

Related Issues

• Closes #...
• Related to #...

Module(s) Affected

• agents
• api
• config
• core
• knowledge
• logging
• services
• tools
• utils
• web (Frontend)
• docs (Documentation)
• scripts
• tests
• Other: ...

Checklist

• [+] I have read and followed the contribution guidelines.
• [+] My code follows the project's coding standards.
• [-] I have run pre-commit run --all-files and fixed any issues.
• [-] I have added relevant tests for my changes.
• [-] I have updated the documentation (if necessary).
• [+] My changes do not introduce any new security vulnerabilities.

Additional Notes

Summary
When using Qwen models via vLLM, the chat template throws a jinja2.exceptions.TemplateError because multiple system messages are being sent to the LLM. Qwen's chat template requires exactly one system message at position 0 of the messages array.

Error Message
jinja2.exceptions.TemplateError: System message must be the beginning.
ValueError: System message must be at the beginning.
Full traceback shows the error originates from:

vllm/renderers/hf.py:502 - safe_apply_chat_template()
transformers/tokenization_utils_base.py:3063 - apply_chat_template()
Qwen's Jinja2 chat template at line 85
Reproduction Steps

Start DeepTutor with Qwen model via vLLM
Open a new chat session
Send any message (e.g., "привет" / "hello")
Observe the error in vLLM logs
Note: This happens even in a brand new chat with no conversation history, if memory files (PROFILE.md/SUMMARY.md) exist.

Root Cause Analysis
Message Construction Flow
The issue occurs in the message construction pipeline where multiple system messages are added:

┌─────────────────────────────────────────────────────────┐
│ 1. _build_messages() in agentic_pipeline.py │
│ ├─ System prompt → {"role": "system", ...} │
│ ├─ Memory context → {"role": "system", ...} ← BUG │
│ └─ Conversation history (may include system) ← BUG │
└─────────────────────────────────────────────────────────┘
↓
┌─────────────────────────────────────────────────────────┐
│ Messages sent to vLLM: │
│ [ │
│ {"role": "system", "content": "thinking prompt"}, │
│ {"role": "system", "content": "## Background Memory"},│
│ {"role": "user", "content": "привет"} │
│ ] │
└─────────────────────────────────────────────────────────┘
↓
┌─────────────────────────────────────────────────────────┐
│ Qwen template error: System must be at position 0 only │
└─────────────────────────────────────────────────────────┘

[OlegSob-glitch]

these edits are taken into account in #295