feat: add hippo-embed workflow + recurring embed step to daily-hippo-learn

[Copilot]

The Hippo memory store has ~490 memories but fewer than 1% are vector-embedded, making semantic recall effectively non-functional. Root cause: hippo embed (and its required dep @xenova/transformers) was never wired into any workflow.

Changes

• New hippo-embed.md — workflow_dispatch-only maintenance workflow for the one-time fix, running on aw-gpu-runner-T4:

  1. hippo audit --fix — prunes flagged junk entries before indexing
  2. npm install -g @xenova/transformers — installs the embedding backend as a CI step (before the agent runs)
  3. hippo embed — generates vectors for all memories
  4. hippo status — reports coverage delta

• daily-hippo-learn.md — adds Step 2.5 after hippo sleep to run hippo embed on every daily cycle, keeping the vector index current as new memories accumulate:

  mcpscripts-hippo args: "embed"
  

  This is incremental — only unembedded memories are processed, so daily overhead is minimal.

[pelikhan]

@copilot install transformers as a step. Use GPU runner (same as daily-news)

[Copilot]

@copilot install transformers as a step. Use GPU runner (same as daily-news)

Done in 9445fe0. @xenova/transformers is now installed in a steps: block so it runs as a CI step before the agent. Also added runs-on: aw-gpu-runner-T4.

[github-actions]

Hey @Copilot 👋 — great work tackling the vector-embedding gap in the Hippo memory store! Fixing semantic recall by wiring up hippo embed into both a one-time maintenance workflow and the daily cycle is a solid, well-scoped change.

One thing worth addressing before this moves out of draft:

• No test coverage — the new hippo-embed.md workflow and the daily-hippo-learn.md change don't appear to have any corresponding test files. Even a basic smoke test or a unit test validating the workflow spec structure would help CI catch regressions.

If you'd like a hand, here's a prompt to get started:

Add unit or integration tests for the hippo embed workflow changes introduced in this PR:
1. Write a test that validates the structure/schema of `.github/workflows/hippo-embed.md` (e.g., required steps are present, the workflow_dispatch trigger is set, the correct runner is specified).
2. Write a test that validates the new Step 2.5 in `.github/workflows/daily-hippo-learn.md` is present and calls `hippo embed` correctly.
Place tests in the appropriate test directory following the existing test conventions for this repo.

Generated by Contribution Check · ● 1.6M · ◷

[Copilot]

Pull request overview

Adds a Hippo embedding maintenance workflow and introduces a recurring embedding step to keep the Hippo memory store’s vector index up to date for semantic recall.

Changes:

• Introduces a new Hippo Embed workflow (workflow_dispatch only) that audits/prunes low-quality entries and embeds all memories.
• Updates the daily Hippo learning workflow instructions to run hippo embed after each hippo sleep.
• Adds the new workflow to the agent-factory status documentation table.

Show a summary per file

File	Description
docs/src/content/docs/agent-factory-status.mdx	Adds the Hippo Embed workflow to the published workflow/status list.
.github/workflows/hippo-embed.md	New maintenance workflow definition and agent instructions for audit + embed + status reporting.
.github/workflows/hippo-embed.lock.yml	Compiled/locked workflow artifact for hippo-embed.md.
.github/workflows/daily-hippo-learn.md	Adds an “embed” step after the daily “sleep” consolidation cycle.

Copilot's findings

Tip

Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

• Files reviewed: 4/4 changed files
• Comments generated: 2

[Copilot]

The new mcpscripts-hippo args: "embed" step will fail in this workflow because @xenova/transformers (required by hippo embed, per the PR description) is never installed here. Add a setup steps: entry (similar to hippo-embed.md) to install the dependency, and re-run the workflow compiler so the change is reflected in daily-hippo-learn.lock.yml (the lock file is what GitHub Actions executes).

[Copilot]

@xenova/transformers is installed here, but the same dependency is also needed by the recurring embed step added to daily-hippo-learn.md. To avoid duplicating setup across workflows (and to ensure future Hippo workflows don’t forget it), consider moving this install into a shared import (e.g., extend shared/hippo-memory.md or add a new shared hippo-embeddings.md import) and reusing it in both workflows.