PlatoAI

A philosophy dialogue app with five conversation modes, retrieval from primary texts, and a source panel for checking passages.

Project at a glance

My responsibility: Built the application solo, including text embeddings, retrieval, dialogue prompts, speaker coordination, and the Next.js interface.
AI / RAG: I built this to test whether retrieval and prompts could keep a dialogue focused on a philosophical method and its source texts. I wanted the speakers to examine arguments through questions while staying consistent with the supplied passages.

Grounding: Cited source texts
Interaction: 5 dialogue modes
Stack: Next.js + Pinecone

Next.jsTypeScriptOpenAIRAGVector DB

Visit project View resume

Product screens

Dialogue Home

Suggested questions and dialogue modes lead straight into a conversation.

Socratic Dialogue

Plato and Socrates debate a modern dilemma in distinct streamed voices.

Source Citations

Each response is grounded in retrieved passages with relevancy scores.

The Modes of Dialogue

Five modes change the interaction contract, from Elenchus to Symposium.

Source Citations

Each response is grounded in retrieved passages with relevancy scores.

Dialogue Home

Suggested questions and dialogue modes lead straight into a conversation.

Socratic Dialogue

Plato and Socrates debate a modern dilemma in distinct streamed voices.

The Modes of Dialogue

Five modes change the interaction contract, from Elenchus to Symposium.

The work

Problem, implementation, and result.

Problem

I wanted to explore whether interactive dialogue could help readers examine philosophical arguments and refer back to the original texts.

Problem

Build

I built dialogue modes that prompt Plato, Socrates, and other speakers to question and debate. The application retrieves relevant passages from indexed texts and displays them in a Source Texts panel.

System design

Outcome

The application combines text ingestion, retrieval, dialogue prompting, and source inspection. The source panel shows the retrieved context without claiming that every answer faithfully represents it.

Delivered behavior

Framing

Scope and constraints.

I built this to test whether retrieval and prompts could keep a dialogue focused on a philosophical method and its source texts. I wanted the speakers to examine arguments through questions while staying consistent with the supplied passages. I split the texts into passages that preserve enough context to follow an argument. Each dialogue mode uses its own prompts with the same retrieval system. Readers can open the Source Texts panel beside the conversation.

Technologies

Next.js

TypeScript

OpenAI

RAG

Vector DB

Key features in scope

Five dialogue modes - Classic Dialogue, The Elenchus, The Symposium, Devil's Advocate, and Modern Dilemmas

Source citations with relevancy scores and readable transcripts for every dialogue

Multi-speaker streamed dialogues between Plato, Socrates, and guest figures

Saved dialogue library, curated featured dialogues, and shareable conversations

Responsibility and context

Role: Built the application solo, including text embeddings, retrieval, dialogue prompts, speaker coordination, and the Next.js interface.

Category: AI / RAG

Engineering

Architecture and decisions.

Technologies

React

Next.js

TypeScript

Tailwind CSS

Framer Motion

Next.js API routes

Node.js

Pinecone

Architecture choices

Embedding pipeline that chunks and vectorizes the works of Plato and Xenophon into Pinecone

Retrieval that scores passages for each prompt and supplies them as dialogue context, with a Source Texts panel for readers to inspect them

Multi-speaker orchestration that keeps Plato, Socrates, and Symposium guests in distinct voices within one streamed response

Five dialogue modes implemented as distinct prompting strategies over the same retrieval core

Next.js App Router frontend with streaming responses, a saved dialogue library, and shareable dialogues

Key decisions

Prompt for questions and argument rather than only explanatory answers

Retrieve passages from the indexed works for each dialogue context

Expose source passages and relevance scores in the interface

Use distinct prompts for each dialogue mode over the same retrieval system

Results

Results and limitations.

Delivered behavior

Built five dialogue modes with streamed responses, a saved dialogue library, and sharing

Added a Source Texts panel with retrieved passages and relevance scores

Connected multi-speaker prompting to a shared retrieval pipeline

Constraints

Keep the dialogue focused on the source material

Distinguish multiple speakers within a streamed response

Keep retrieval responsive enough for conversational use

Lessons

What the build taught me.

Chunks need enough context to preserve a philosophical argument

Dialogue method and speaker voice need separate attention in prompt design

Source inspection should be part of the reading experience

More case studies

AI / Workflow Tooling

Airtable Base Doctor

An Airtable review app that checks schema and sample records, then generates interface previews, redesign proposals, and migration plans.

Read case study

Digital Platform / AI

Dr. Anne Hussain Platform

A book, blog, and podcast platform with an assistant that retrieves published passages and presents source citations.

Read case study

AI / Social Product

TweetNet

A social simulation where configurable AI personas write posts and respond to one another in a shared feed.

Read case study

SaaS / Productivity

CamelCollect.io

A web app and browser extensions for saving LinkedIn posts into organized collections, with onboarding and subscription billing.

Read case study