Python: issue-5260 adds valkey message history

[MatthiasHowellYopp]

Motivation and Context

Fixes #5260

The current agent-framework-redis package depends on redisvl, which requires Redis Stack's RediSearch module. This creates an incompatibility for teams running Valkey — whether self-hosted or through managed cloud services (AWS ElastiCache, GCP Memorystore) — since RedisVL has known incompatibilities with Valkey's search module. Additionally, even the RedisChatMessageStore (which only needs basic key-value operations) pulls in RedisVL as a transitive dependency.

This PR adds a dedicated agent-framework-valkey package that uses valkey-glide (the official Valkey Python client) directly, with no RedisVL dependency.

Description

New package: agent-framework-valkey (python/packages/valkey/)

Two components:

ValkeyChatMessageStore — Persistent chat message storage implementing the HistoryProvider protocol. Uses only basic Valkey key-value/list operations via valkey-glide, so it works with any Valkey (or Redis OSS) server with no search module required.

ValkeyContextProvider — Long-term memory context provider implementing the ContextProvider protocol. Uses Valkey's native FT.CREATE / FT.SEARCH commands via valkey-glide's custom_command API for full-text and optional hybrid vector search. Requires valkey-search >= 1.2 (ships with valkey-bundle >= 9.1.0).

The package follows the same structure and patterns as agent-framework-redis and agent-framework-mem0. Key design decisions:

Uses valkey-glide (official Valkey client, Apache-2.0 licensed) instead of RedisVL
Vector search uses Valkey's native FT commands directly rather than an abstraction layer, keeping the dependency tree minimal
embed_fn is typed as Callable[[str], Awaitable[list[float]]] — a proper async callable type rather than a framework-specific vectorizer class, making it easy to plug in any embedding provider
numpy is an optional dependency under the [vector] extra, so users who only need ValkeyChatMessageStore don't pull it in
Includes a sample (
valkey_sample.py
) demonstrating both components with Bedrock
Changes:

python/packages/valkey/ — new package with implementation, tests, README, LICENSE
pyproject.toml
— added workspace source and pyright test environment for valkey
PACKAGE_STATUS.md
— registered as alpha
AGENTS.md
— added valkey to Storage & Memory section
uv.lock
— updated with valkey-glide resolution
All checks pass: ruff formatting/linting, pyright, mypy, and 51 unit tests at 84% coverage.

Contribution Checklist

• The code builds clean without any errors or warnings
• The PR follows the Contribution Guidelines
• All unit tests pass, and I have added new tests where possible
• Is this a breaking change? If yes, add "[BREAKING]" prefix to the title of the PR.

[MatthiasHowellYopp]

@microsoft-github-policy-service agree [company="Amazon"]

[MatthiasHowellYopp]

@microsoft-github-policy-service agree company="Amazon"

[MatthiasHowellYopp]

@moonbox3 - I'm getting this error:
Run actions/labeler@v6
with:
configuration-path: .github/labeler.yml
sync-labels: false
dot: true
Error: Error: Parameter token or opts.auth is required
Error: Parameter token or opts.auth is required

Is there something that can be done?