Botting up Agent taking so much time on self hosting.

[Morris-Zin]

Please help I really want to use Agent feature..

### khoj-server-1

khoj-sandbox-1

khoj-computer

khoj-search-1

This recently happens too after uploading documents.

[yang1002378395-cmyk]

需要更多信息

从截图看有 3 个容器在运行：

• khoj-server-1
• khoj-sandbox-1
• khoj-search-1

请提供以下信息帮助诊断：

1. 启动时间：Agent 启动大概需要多久？（5分钟？10分钟？）

2. 日志输出：运行 docker-compose logs khoj-server-1 | tail -50 查看启动日志

3. 硬件配置：

   • CPU 核心数
   • 内存大小
   • 是否使用 GPU？

4. 模型配置：

   • 使用哪个模型？（GPT-4、Claude、本地模型？）
   • KHOJ_CHAT_MODEL 环境变量设置

常见慢启动原因：

• 本地模型首次下载（几GB）
• 向量索引构建（大量文档时）
• Docker 资源限制
• 网络问题（模型下载慢）

提供日志我可以帮你定位具体瓶颈。

[diamond-one]

I’m experiencing the same issue. Here is updated diagnostic information with runtime logs:

---

Containers running:

• khoj-server-1
• khoj-sandbox-1
• khoj-search-1

---

Startup time:

• Server startup: ~30–60 seconds
• Chat works normally
• Agent creation: hangs indefinitely (spinner continues, no response)

---

Log output (server):
Captured using docker compose logs -f server while attempting to create an agent:

[19:43:57] Initializing Khoj v1.42.10
[19:43:57] No migrations to apply
[19:43:57] Chat model options updated
[19:43:57] Starting Khoj
[19:43:57] Schedule Leader elected
[19:43:57] Started Background Scheduler

[19:43:59] Loaded embedding model thenlper/gte-small (~1.9s)
[19:44:00] Loaded cross-encoder model mixedbread-ai/mxbai-rerank-xsmall-v1 (~0.6s)

[19:44:00] Khoj is ready to engage
[19:44:00] Uvicorn running on http://0.0.0.0:42110

Important observation:

• After clicking “Create Agent”, no additional logs are produced
• No HTTP requests, errors, or backend activity appear
• Server remains completely idle

---

Hardware configuration:

• CPU: Intel i5-13400F (16 cores)
• RAM: 31 GB (≈9 GB available)
• GPU: NVIDIA RTX 4060 Ti (16 GB VRAM)

---

Model configuration:

• Local model via Ollama (no external API keys)

• Model: llama3:8b

• GPU acceleration confirmed (100% usage during chat)

• API type: OpenAI-compatible

• Base URL:

http://<local-ip>:11434/v1/

Environment:

OPENAI_BASE_URL=http://<local-ip>:11434/v1/
KHOJ_DEFAULT_CHAT_MODEL=llama3:8b

---

Connectivity verification:

• Container can reach Ollama:

curl http://<local-ip>:11434/v1/models

→ returns valid model list

• Chat successfully invokes model (ollama ps shows active usage)

---

Observed behavior:

• Chat works correctly

• Ollama integration is confirmed working

• Creating an agent:

  • UI spinner continues indefinitely
  • No backend logs are generated
  • No model activity occurs

---

Key finding:
There appears to be no request reaching the backend when creating an agent, as:

• no logs are emitted
• no API calls are triggered
• no model activity occurs

---

Summary:
The system is fully functional for chat and inference, but agent creation:

• does not trigger any backend request
• produces no logs
• does not invoke the configured model

This suggests a frontend issue or a failure before request dispatch to the backend.

---

Happy to provide browser console/network logs if helpful.

[diamond-one]

I've tried again today - seems to create agents.
I can't explain why it's behaving differently ( I didn't even restart my computer since yesterday) but it's working now