agent-discover 1.3.8 → 1.3.9
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/CHANGELOG.md +8 -1
- package/README.md +3 -1
- package/agent-desk-plugin.json +1 -1
- package/package.json +1 -1
package/CHANGELOG.md
CHANGED
|
@@ -5,6 +5,13 @@ All notable changes to this project will be documented in this file.
|
|
|
5
5
|
The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
|
|
6
6
|
and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
|
|
7
7
|
|
|
8
|
+
## [1.3.9] - 2026-04-15
|
|
9
|
+
|
|
10
|
+
### Changed
|
|
11
|
+
|
|
12
|
+
- **Bench README rewritten to reflect reproducible current-state numbers.** Headline replaced with the scaling-of-first-turn-tokens table at N ∈ {10, 100, 1000, 3000} on OpenCode + gpt-5-mini: discover flat at ~20.8k, eager linear (20.9k → 32.4k → 160.9k → context overflow at N=3000). Adversarial CRUD pack now shows both arms at 100% accuracy; discover slightly more expensive on end-to-end multi-turn cost ($0.086 vs $0.068) due to `find_tool` output accumulation in conversation history. Added an explicit multi-turn cost caveat section.
|
|
13
|
+
- **Top-level README elevates the no-session-restart differentiator.** Prominent callout directly under the intro noting that every MCP client today — including Claude Code — requires a full restart to pick up a newly registered MCP server, and agent-discover is the only path to register a new server and have it become discoverable in the same session.
|
|
14
|
+
|
|
8
15
|
## [1.3.2] - 2026-04-12
|
|
9
16
|
|
|
10
17
|
### Changed
|
|
@@ -107,7 +114,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
|
|
|
107
114
|
- **FTS5 + BM25 ranking on `server_tools`** (migration v4). Adds a `server_tools_fts` virtual table with name × 4 / description × 1 column weighting, plus a query preprocessor that expands verb synonyms (`fetch → get`, `cancel → delete`, etc.) and singularizes plurals (`subscriptions → subscription`).
|
|
108
115
|
- **Embedding columns on `server_tools`** (migration v5). `embedding` (TEXT, base64-encoded float32) and `embedding_model` columns for semantic search. Embeddings are optional; tools without one fall back to BM25 ranking only.
|
|
109
116
|
- **Hybrid retrieval (`searchToolsHybrid`).** Brute-force cosine similarity over the entire embedded catalog + BM25 candidate union, scored 70% semantic / 30% lexical. Closes the natural-language gap that pure BM25 misses (e.g. "billing arrangement" → "subscription").
|
|
110
|
-
- **Bench harness** under `bench/` comparing eager tool loading vs deferred discovery. Real Claude Code (`bench/drivers/cli.ts`) and OpenCode (`bench/drivers/opencode.ts`) drivers, isolated bench DB, scoring with `success` / `choice_accuracy` / `distractor_call_rate` / `refusal_rate` metrics, and a standalone `bench/rescore.ts` that re-applies the current scoring logic to captured event streams without spending fresh API tokens. Headline result
|
|
117
|
+
- **Bench harness** under `bench/` comparing eager tool loading vs deferred discovery. Real Claude Code (`bench/drivers/cli.ts`) and OpenCode (`bench/drivers/opencode.ts`) drivers, isolated bench DB, scoring with `success` / `choice_accuracy` / `distractor_call_rate` / `refusal_rate` metrics, and a standalone `bench/rescore.ts` that re-applies the current scoring logic to captured event streams without spending fresh API tokens. Headline reproducible result on OpenCode + gpt-5-mini: discover's first-turn input tokens are flat at ~20.8k across N ∈ {10, 100, 1000, 3000}; eager's grow linearly (20.9k → 32.4k → 160.9k → context overflow at N=3000, where eager can no longer complete the task). Full write-up in `bench/README.md`.
|
|
111
118
|
|
|
112
119
|
## [1.1.4] - 2026-04-09
|
|
113
120
|
|
package/README.md
CHANGED
|
@@ -9,6 +9,8 @@
|
|
|
9
9
|
|
|
10
10
|
**MCP server registry and marketplace.** Discover, install, activate, and manage MCP tools on demand. Acts as a dynamic proxy -- activated servers have their tools merged into the registry's own tool list, so agents can use them without restarting.
|
|
11
11
|
|
|
12
|
+
> **Every MCP client today — Claude Code, Cursor, Codex CLI, Aider, Continue, plain MCP clients — requires a full agent-session restart to pick up a newly registered MCP server.** The tool catalog is frozen at startup. agent-discover is the only path to register a new server and have it become discoverable in the same running session. This is the one differentiator that survives against every host, even those with their own built-in deferred-tool loaders.
|
|
13
|
+
|
|
12
14
|
Search spans the **official MCP registry**, **npm**, and **PyPI** in one query, so popular servers that aren't in the official index (Microsoft `@playwright/mcp`, `@modelcontextprotocol/server-*`, `mcp-server-fetch`, `mcp-server-git`, …) all show up.
|
|
13
15
|
|
|
14
16
|
Built for AI coding agents (Claude Code, Codex CLI, Gemini CLI, Aider) but works equally well with any MCP client, REST consumer, or WebSocket listener.
|
|
@@ -60,7 +62,7 @@ Static MCP configs mean every server is always running, even when unused. Adding
|
|
|
60
62
|
- **Real-time dashboard** -- web UI at http://localhost:3424 with Servers and Browse tabs, dark/light theme, WebSocket updates
|
|
61
63
|
- **3 transport layers** -- MCP (stdio), REST API (HTTP), WebSocket (real-time events)
|
|
62
64
|
- **Declarative setup file** -- set `AGENT_DISCOVER_SETUP_FILE` to a JSON file listing servers to ensure-registered on startup. Idempotent (skips existing). Supports `auto_activate`, env var secret refs (`$VAR`), and tags. Automatically also reads a `.local.json` variant (e.g. `discover-setup.local.json`) for machine-specific servers with secrets. New `registry({ action: "sync" })` MCP action and `POST /api/sync` REST endpoint for on-demand re-read.
|
|
63
|
-
- **Bench harness** -- under `bench/`, comparing eager tool loading vs deferred discovery against real
|
|
65
|
+
- **Bench harness** -- under `bench/`, comparing eager tool loading vs deferred discovery against real OpenCode + gpt-5-mini. Reproducible structural result: discover's first-turn input tokens are flat in N (~20.8k across N ∈ {10, 100, 1000, 3000}); eager's grow linearly (20.9k → 32.4k → 160.9k → context overflow at N=3000). End-to-end accuracy and multi-turn cost numbers are noisier and model-dependent — see [`bench/README.md`](bench/README.md) for what reproduces and what doesn't.
|
|
64
66
|
|
|
65
67
|
---
|
|
66
68
|
|
package/agent-desk-plugin.json
CHANGED
package/package.json
CHANGED