Skip to main content
Glama
senih25

Recall

by senih25
OverviewSchemaSourceRelated ServersDiscussionsAdmin
Python
Remote

Recall — a hands-free spaced-repetition learning coach for Alexa+

License: MIT

Turn what you read into voice-quizzed mastery. Feed Recall your notes, docs, or a topic; it builds flashcards and coaches you hands-free through Alexa+ on an SM-2 spaced-repetition schedule.

Built for the Build, Ship, Shape: Amazon Developer Hackathon — Alexa+ track, with the AWS Builder and Open Source mini challenges.

How it fits the tracks

Requirement

How Recall meets it

Alexa+ track — self-hosted MCP server, spec 2025-11-25+, Streamable HTTP

recall/server.py runs a FastMCP server on streamable-http; initialize negotiates protocol 2025-11-25.

AWS Builder mini

recall/bedrock.py calls Amazon Bedrock (Claude, Converse API) for flashcard generation and free-text answer grading.

Open Source mini

MIT-licensed, public repo, this contribution.

Related MCP server: Flashcard MCP

Architecture

Recall architecture

Alexa+  ──MCP (Streamable HTTP)──►  recall.server (FastMCP)
                                       │
                 ┌─────────────────────┼─────────────────────┐
                 ▼                     ▼                       ▼
          bedrock.py             srs.py (SM-2)            store.py (JSON)
      Bedrock Converse:     next interval / easiness    per-user cards,
      make + grade cards      from recall grade          due tracking

The SM-2 scheduler is real logic, not an LLM wrapper: recall grades (0–5) drive the easiness factor and interval so hard cards resurface and mastered cards fade out. Bedrock supplies the language understanding (turning source text into cards, grading a spoken answer); if AWS credentials are absent the server falls back to a deterministic heuristic so it still runs locally.

MCP tools

Tool

Purpose (spoken-first responses)

add_source(text, topic, count)

Generate flashcards from a chunk of text.

next_review()

Get the next due card's question to read aloud.

grade(card_id, answer)

Grade a spoken answer, reschedule via SM-2.

progress()

Speak a recap: total / learned / due.

Run locally

pip install -r requirements.txt
python -m recall.server          # MCP: Streamable HTTP on http://127.0.0.1:8000/mcp
python -m recall.webapp          # Alexa+ voice simulator on http://127.0.0.1:8080

The MCP server is the Alexa+ track integration; the web app is the permitted Alexa+ experience simulator (browser Web Speech API). Both call the same recall/core.py logic. See docs/ for the Alexa+ integration guide, demo script, friction log, and submission draft.

Smoke test the transport:

curl -s -X POST http://127.0.0.1:8000/mcp \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-11-25","capabilities":{},"clientInfo":{"name":"smoke","version":"0"}}}'

AWS Bedrock (AWS Builder mini)

Set credentials and a model id; the server uses the Bedrock Converse API:

export AWS_REGION=us-east-1
export BEDROCK_MODEL_ID=anthropic.claude-3-5-sonnet-20241022-v2:0
# standard AWS credential chain (env, profile, or role)

Tests

python recall/srs.py         # SM-2 scheduler self-check
python recall/store.py       # persistence self-check
python recall/bedrock.py     # offline grading/generation fallback
PYTHONPATH=. python tests/test_flow.py   # end-to-end add → review → grade

License

MIT — see LICENSE.

This server cannot be deployed

One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

Maintenance

ActivityMaintained
ResponsivenessNo issues

Resources

Related MCP Connectors

Related MCP Servers