The Code LawyersIndependent thinking. Useful software.Our work

Built by The Code Lawyers · AI assistants & workflow automation

LampAn independent AI assistant, also integrated into Colyf.

  • Independent product
  • Integrated into Colyf
  • Web & mobile in Colyf
Explore LampUse Lamp independently
  • TypeScript
  • Python
  • LLM APIs
  • PostgreSQL
  • React / React Native
  • Scoped tool integrations

Useful context. Questions before action. Changes you review and confirm.

The project

What is Lamp?

Lamp is an independent AI assistant product by The Code Lawyers, available outside Colyf and also integrated into Colyf. It connects conversation with useful context and supported workflows, asking questions when a request needs more detail and keeping people involved when a change is proposed.

Inside Colyf, Lamp is available through the web and mobile experiences. It works with the signed-in user's institution, role and permitted resources. Supported reads can inform a response; supported writes are presented as reviewable actions before confirmation.

The wider project also includes a separate workspace agent that works in Telegram, WhatsApp and Slack. That workspace and the Colyf integration have separate identities, storage and tool boundaries. A Colyf conversation does not grant access to the workspace agent's Jira or host tools.

Designed around real work

What Lamp brings together.

01

Context with boundaries

In Colyf, requests use the signed-in identity and institution scope, so answers and available tools reflect the person's actual access.

02

Questions before action

Lamp can ask follow-up questions and offer choices when details are missing instead of silently guessing required fields.

03

Review and confirm

Supported changes are shown for review and require explicit confirmation before execution through normal application routes.

04

Private saved chats

Colyf conversations persist across sessions and are isolated by user and institution. An administrator does not gain another user's private chat history.

05

Document context

Supported, accessible attachments can inform a conversation within the application's file validation and access rules.

06

Separate integrations

The Colyf assistant and the workspace agent in Telegram, WhatsApp and Slack keep their identities, memories and tool permissions separate.

How Lamp works inside Colyf

From a question to a confirmed change.

Illustrative example. Available actions depend on the signed-in user, their role, course scope and enabled features.

Lampin ColyfIllustrative
Design-brief.pdfSelected documentCreate a research journal assignment for Human-centred design, due Friday.
I can prepare that from the brief. Which class should receive it?Studio AStudio B
Proposed changeConfirmed
Action
Create assignment
Subject
Human-centred design · Studio B
Title
Design research journal
Due
Friday, 11:59 PM
or type yes / cancelCancelConfirm
  • Membership and role re-checked
  • Course scope confirmed for Studio B
  • Sent through Colyf's assignment route
Ask Lamp…
  1. 01 · Context

    Start with your context.

    Ask in everyday language on the web or in the mobile app. Attach an approved PDF, image or text file when it helps the request.

  2. 02 · Clarify

    Questions before action.

    When a required detail is missing, Lamp asks and offers selectable replies instead of silently guessing.

  3. 03 · Review

    See the proposal first.

    A review card shows the actual fields that would change. It belongs to one user and institution, and it expires if it is left waiting.

  4. 04 · Confirm

    Move forward with control.

    Confirm or cancel with a button, or by typing yes or cancel. When the action runs, Colyf checks your current role and permissions again.

Architecture in motion

Every request follows the same guarded path.

Identity and scope come from authentication, a closed tool catalog limits what can be proposed, and confirmation re-enters the application's own routes.

Lamp request flow: you ask on web or mobile; Lamp gathers permitted context; a closed tool catalog limits actions to your role and enabled features; Lamp prepares a reviewable proposal; you confirm; application routes check access again; records and private threads are stored in PostgreSQL. The workspace agent for Telegram, WhatsApp and Slack is isolated with separate identity and tools.isolatedsaved chatsYou askWeb or mobileLampConversation + contextClosed tool catalogRole + enabled featuresReviewable proposalTied to user + institutionYou confirmExplicit, every changeApplication routesAccess checked againPostgreSQLRecords + private threadsWorkspace agentTelegram, WhatsApp, SlackLamp request flow: you ask on web or mobile; Lamp gathers permitted context; a closed tool catalog limits actions to your role and enabled features; Lamp prepares a reviewable proposal; you confirm; application routes check access again; records and private threads are stored in PostgreSQL. The workspace agent for Telegram, WhatsApp and Slack is isolated with separate identity and tools.isolatedsaved chatsYou askWeb or mobileLampConversation + contextClosed tool catalogRole + enabled featuresReviewable proposalTied to user + institutionYou confirmExplicit, every changeApplication routesAccess checked againPostgreSQLRecords + private threadsWorkspace agentTelegram, WhatsApp, Slack
  • Authenticated identity and a closed tool catalog
  • User-confirmed proposals with normal application authorization
  • Separate private memory and integration boundaries

Inside the build

Explore the engineering.

01 / 03

From Conversation to Useful Work

Give an assistant useful context and supported actions without bypassing the application's permissions.

Read the case study 02 / 03

Lamp Context & Action Architecture

Keep identity, conversation memory and tool authority separate while connecting the assistant to application workflows.

Read the case study 03 / 03

Lamp Technology Stack

Application-scoped AI orchestration, durable conversation storage and separate workspace integrations.

Read the case study

A few useful answers

About Lamp.

What is Lamp AI?

Lamp is an independent AI assistant product by The Code Lawyers. It is also integrated into Colyf, where it helps people find authorized institution information and prepare supported changes for review on web and mobile.

Can I use Lamp without Colyf?

Yes. Lamp is available as an independent product as well as an integrated assistant in Colyf. Contact The Code Lawyers to discuss using Lamp outside Colyf.

Can Lamp make changes automatically?

Within Colyf, supported writes require explicit review and confirmation. Execution goes through the normal authenticated application routes, which check the user's current permissions again.

Can Lamp access every school record?

No. Available context and tools depend on the signed-in user's institution, role, course and resource access, plus enabled features. Conversation text does not expand those permissions.

Is the workspace agent in Telegram, WhatsApp or Slack connected to my Colyf chat history?

They are separate integrations. Colyf's private chats and application permissions do not expose the workspace agent's identities, conversations, Jira integration or host tools.

Connected projects

Connected learning and campus operations, with Lamp AI integrated.

Explore Colyf
Colyf product illustration

Your next project

Build something that
makes work clearer.

Talk to The Code Lawyers
AI assistantsCustom softwareAll projects