Skip to main content

Build Status Coverage

echo-words

From a word you meet to a word you remember.

Understand words and expressions in context, hear their pronunciation, and send audio-backed flashcards straight to AnkiWeb — ready to sync into the Anki apps you already use.

echo-words explains; Anki makes you remember.

A word, a phrase, a sentence
Analysis, audio, card in Anki
A sentence, translated and explained
  • Look up a word. Type a foreign word, or one in your own language to find the foreign one. Explore meanings, usage, and translated examples. Choose the sense you want to learn.
  • Learn from a sentence. Get a translation and an explanation of the difficult parts, then tap a word or expression to explore it.
  • Keep it in Anki. Each selected sense becomes four cards covering recognition and production, with and without context, in its source-language deck.

Documentation · Install it by hand, or let Claude or ChatGPT do it for you.

Under the hood

Getting a complete answer from a changing model pool. Free LLM providers vary in availability, speed, and instruction-following. Through llmbroker, echo-words races two models and selects the first complete, usable answer. Streaming lets the reader start earlier; recovery preserves an existing explanation if card creation fails.

If your application needs a pool of free LLM providers with automatic failover and streaming, llmbroker is available as a standalone Python library.

Checking what the cards actually teach. Valid JSON can still contain the wrong meaning, an invented origin, or a word from the wrong language. Model-facing changes go through real-model benchmarks and a separate agent's review of the concrete answers. The evaluation records document both findings and remaining limitations.

Keeping the backend operational in less than 1G. FastAPI serves the Vue PWA and maintains Anki through its headless Python library, syncing directly with AnkiWeb. There is no separate application database or running Anki desktop instance. Access to the web app is restricted to a Tailscale network.

Development process

In this project, I experimented with a staged AI development workflow: Fable for design, sol for the initial implementation with iterative reviews automatically coordinated by Astra, and Opus for subsequent fixes, coordinating its own review cycles.

The continuity between those stages lives in the repository: a functional specification, decision records, and working plans. Agent instructions define the verification gates, including browser tests and real-model evaluation for changes to prompts or answer handling.

Contributing
uv sync
npm --prefix webapp ci
uv run playwright install chromium webkit
uv run inv dev     # http://127.0.0.1:8080
uv run inv pre     # lint, format, type-check, file hygiene
uv run inv test    # Python + frontend suites

See Development and Deploy to Oracle Cloud in the docs.

Allure test report

Metadata

Release files for echo-words 1.15.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for echo-words 1.15.0
File Size Uploaded Provenance Actions
echo_words-1.15.0.tar.gz 2.1 MB Without provenance Details for echo_words-1.15.0.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for echo-words 1.15.0
File Interpreter ABI Platform Actions
echo_words-1.15.0-py3-none-any.whl Python 3 none any Details for echo_words-1.15.0-py3-none-any.whl

Total release size: 2.2 MB

Release files / echo_words-1.15.0.tar.gz

Download URL echo_words-1.15.0.tar.gz
Size 2.1 MB
Tags Source
SHA-256 checksum
How to use checksums
76c9c0a741d2c79eb723eda491c9f2d5a1f5b996c166fb7e697d7e2bee6a63a7
BLAKE2b-256 checksum
How to use checksums
716be947de28f1bcb95fd5899375b92d3294800a8a128636a30d152bda9be8c4
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.13.16

Release files / echo_words-1.15.0-py3-none-any.whl

Download URL echo_words-1.15.0-py3-none-any.whl
Size 108.0 kB
Tags Python 3
SHA-256 checksum
How to use checksums
9b0e13b5d2be06e7d96bb1b7cdb36532d78b19d05f23e4e590441b4c8bb5a495
BLAKE2b-256 checksum
How to use checksums
513a8376918197e7df2b4723fb74a62dbb7200f487f4dcf9eaa8f0efd4fd7924
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.13.16

Security for echo-words 1.15.0

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Release history Release notifications | RSS feed

This release

1.15.0 This release

2 release files No provenance

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page