# Ollive Inference Chatbot: An LLM Chat System with Inference Logging

> Ollive is a full-stack LLM chatbot that includes a lightweight inference logging SDK, a near-real-time ingestion API, and PostgreSQL storage. It supports multiple providers (Gemini, OpenAI, Anthropic), streaming responses, and a real-time metrics dashboard.

- 板块: [Openclaw Llm](https://www.zingnex.cn/en/forum/board/openclaw-llm)
- 发布时间: 2026-05-23T09:12:18.000Z
- 最近活动: 2026-05-23T09:23:50.678Z
- 热度: 159.8
- 关键词: LLM, 聊天机器人, 推理日志, 监控, PostgreSQL, 多提供商, 流式响应, PII脱敏
- 页面链接: https://www.zingnex.cn/en/forum/thread/ollive-inference-chatbot-llm
- Canonical: https://www.zingnex.cn/forum/thread/ollive-inference-chatbot-llm
- Markdown 来源: floors_fallback

---

## Introduction / Main Floor: Ollive Inference Chatbot: An LLM Chat System with Inference Logging

Ollive is a full-stack LLM chatbot that includes a lightweight inference logging SDK, a near-real-time ingestion API, and PostgreSQL storage. It supports multiple providers (Gemini, OpenAI, Anthropic), streaming responses, and a real-time metrics dashboard.

## Original Author and Source

- **Original Author/Maintainer**: Nightstorm26
- **Source Platform**: GitHub
- **Original Title**: ChatBot (Ollive Inference Chatbot)
- **Original Link**: https://github.com/Nightstorm26/ChatBot
- **Publication Time**: May 23, 2026

## Project Overview

Ollive Inference Chatbot is a full-stack LLM chat application that includes three core components: a lightweight inference logging SDK, a near-real-time ingestion API, and a PostgreSQL database for storing messages and inference metadata.

This project addresses a key need in LLM applications: how to reliably record and monitor inference calls while maintaining low latency and a good developer experience.

## Multi-turn Conversation Support

The system maintains conversation history (latest 20 messages) and sends it to the model. This is implemented via a simple message list instead of complex token-aware context management.

## Multi-provider Support

- Google Gemini (default)
- OpenAI
- Anthropic

Users can switch between different providers and models during a conversation.

## Streaming Responses

Uses SSE (Server-Sent Events) to implement token-by-token streaming responses, providing a better user experience.

## Inference Metrics Dashboard

Real-time 24-hour panel displays:
- Latency statistics
- Throughput
- Error distribution
- Statistics per provider

## PII Redaction

Sensitive information is redacted in log previews:
- Email addresses
- Phone numbers
- SSN
- Bank card numbers
- API keys