chroma-memory

Store and retrieve per-customer conversation history in ChromaDB.

152|46|Updated Mar 23, 2026
One-click install
npx skills add https://github.com/iPythoning/b2b-sdr-agent-template --skill chroma-memory-ipythoning
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: chroma-memory
Source: https://github.com/iPythoning/b2b-sdr-agent-template/tree/main/skills/chroma-memory
Command: npx skills add https://github.com/iPythoning/b2b-sdr-agent-template --skill chroma-memory-ipythoning

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill prevents “anti-amnesia” failures by preserving every conversation turn and by providing a safe recall path for returning customers and daily recovery snapshots.

Core Features & Use Cases

  • Per-customer conversation memory (L3): Stores each turn with strict customer_id isolation so retrieval stays tenant-safe.
  • Semantic search and retrieval: Finds relevant prior discussion across a customer’s stored history for smoother, context-aware responses.
  • Daily CRM snapshot fallback (L4): Saves a lightweight daily snapshot marker as disaster recovery when full history is unavailable.

Use case: When a B2B lead comes back after several days and asks about pricing, objections, or commitments, the agent can retrieve the most relevant past quotes and promises for that same customer_id.

Quick Start

Store a conversation turn by instructing the agent to run: chroma:store --customer "+971501234567" --turn 5 --user "What's the price for 500 units?" --agent "Let me prepare a detailed quote..." --stage qualifying --topic pricing.

Frequently Asked Questions about chroma-memory

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I store and retrieve per-customer conversation history for B2B sales workflows?

Store and retrieve per-customer conversation history in ChromaDB to maintain long-term sales context. It saves each dialogue turn with strict customer_id isolation, ensuring tenant-safe retrieval for returning B2B leads.

Can I automatically tag sales conversations by quotes, objections, and commitments?

Yes, automatic content tagging categorizes stored conversation turns by quotes, commitments, objections, orders, and samples. This allows quick semantic search and retrieval of relevant prior sales discussions for each customer.

What is a CRM snapshot fallback and when do I need it for conversation memory?

A CRM snapshot fallback is a lightweight daily marker saved to ChromaDB for disaster recovery. You need it when full conversation history is unavailable, ensuring a safe recall path for returning customers.

How do I recall recent conversation history when a returning customer asks about pricing?

Recall recent history by running a deterministic CLI command flow with the customer_id. The semantic search finds the most relevant prior quotes, objections, and commitments stored for that specific returning customer.

Does customer conversation memory support customer_id isolation for tenant-safe retrieval?

Yes, customer conversation memory applies strict customer_id-isolated persistence in ChromaDB. This ensures all stored dialogue turns and semantic search results remain completely tenant-safe across different customers.

What are the limitations of using semantic search for sales pipeline conversation memory?

Semantic search for conversation memory requires deterministic CLI command flows for store, search, recall, and stats operations. It relies on ChromaDB persistence and daily CRM snapshot markers as a fallback when full history is unavailable.