prjct-cli 2.38.1 → 2.39.0

package/CHANGELOG.md

@@ -2,6 +2,24 @@
 
 ## [Unreleased]
 
+The long-lived daemon no longer serves stale code, and embeddings reach any provider.
+
+### Fixed
+- **Daemon could serve output from an outdated build (the recurring "stale daemon" trap).** Two compounding defects: staleness was detected *after* the daemon decided to serve, so the request that first observed a new build/install was still answered by the old code; and on refusal the client printed `Daemon restarting — retry the command` and exited 1, so the command never ran (1st call stale → 2nd fails → 3rd fresh). Now staleness is detected *before* serving, the daemon refuses cleanly with a `retry` signal (the request never executes → zero side effects), and **both `bin/prjct.ts` and the daemon shim** (`scripts/build.js` → `dist/bin/prjct.mjs`, what users actually run) transparently fall through to direct in-process execution on the fresh code — no error, no manual re-run. Hooks degrade to the fail-soft `{}` no-op. The global-install version-drift check is now time-throttled (1s) instead of every-10-commands (which could leak up to 9 stale responses). Closes the "stale daemon caches old hook code" gotcha.
+- **`prjct embeddings test` no longer truncates its failure** to 65 chars (`embeddings endpoi…`). It now prints the full endpoint response plus an actionable hint (401 → base-url/key mismatch, 404 → wrong route, network → unreachable) — the one command meant to diagnose connectivity stopped hiding the diagnosis.
+- **Embeddings test isolation:** `global-config` resolves its directory lazily from `PRJCT_CLI_HOME`, so config/embeddings tests no longer read the developer's real `~/.prjct-cli`.
+
+### Added
+- **Zero-config embeddings setup — paste just the key.** `prjct embeddings set --key <k>` now infers the base URL from the key's prefix (`sk-or-…` → OpenRouter, `sk-…` → OpenAI, `pa-…` → Voyage; `sk-ant-`/Groq/unknown → keep default), so `--base-url` is optional for known providers. An explicit `--base-url` still wins, and detection re-fires when you switch keys. `set` is now partial-update friendly too: `set --key …` keeps your existing model/base URL instead of resetting them.
+- **Embeddings support ANY OpenAI-compatible provider, including non-Bearer ones.** Already worked with OpenAI, OpenRouter, Ollama, Together, Mistral, Voyage, Jina, LM Studio, … via `--base-url` + `--model` + `--key`; now providers that deviate from `Authorization: Bearer` are covered too — e.g. **Azure OpenAI** (`api-key` header + `api-version` query) and custom gateways. New `prjct embeddings set` flags: `--auth-header <h>`, `--auth-scheme <s|none>` (`none` = raw key), `--headers "k=v,…"`, `--query <qs>`. Default stays `Bearer`/`authorization`, so existing providers are unchanged. Vector dimensionality is detected from the response (no hardcoded size).
+
+## [2.39.0] - 2026-06-02
+
+### Features
+
+- zero-config embeddings (any provider) + fix daemon never serves stale code (#412)
+
+
 ## [2.38.1] - 2026-06-02
 
 ### Performance

package/README.md

@@ -201,7 +201,7 @@ Cursor / Windsurf use the same commands with a `/` prefix: `/capture`, `/task`,
 | `prjct status <value>` | Inline status change on the active task (`done`, `paused`, `active`, …). |
 | `prjct tag <k:v>` | Tag the active task (`type:bug`, `domain:auth`, …). |
 | `prjct remember <type> "<content>"` | Persist a memory entry (decision, learning, gotcha, …). |
-| `prjct embeddings <set\|status\|test\|clear>` | Configure the global BYOT embeddings provider (one secure key, all projects). |
+| `prjct embeddings <set\|status\|test\|clear>` | Configure the global BYOT embeddings provider — any OpenAI-compatible API (OpenAI, OpenRouter, Ollama, Azure, …), one secure key, all projects. |
 | `prjct ship [name]` | Run the project's ship workflow (commit, push, PR, persist). |
 | `prjct sync` | Re-index files, git co-change, imports; refresh project analysis. |
 | `prjct regen` | Full rebuild of the Obsidian vault snapshot from SQLite. |
@@ -282,7 +282,7 @@ prjct hooks <install|uninstall|status>  Git hooks for auto-sync
 prjct context <memory|learnings|wiki>  Recall memory / sync the vault
 prjct review-risk        Advisory change-size + delivery-geometry hint (read-only)
 prjct workflow ["config"]  Configure hooks via natural language
-prjct stop / restart     Background daemon control
+prjct stop / restart     Background daemon control (self-reloads on stale code; manual restart rarely needed)
 prjct login / logout / auth   Cloud sync authentication
 prjct update             Update CLI system-wide (alias: prjct upgrade)
 prjct --version / --help
@@ -310,15 +310,44 @@ Memory is FTS5-backed (SQLite) and persona-filtered. Recall blends three signals
 
 On by default for **every** project — no setup, no key, no native dependency. A built-in pure-JS embedder (feature-hashed character n-grams) vectorizes memory into SQLite so recall catches morphological / cross-vocabulary matches BM25 misses (`auth`≈`authentication`).
 
-Want higher quality? Bring your own key once, globally:
+Want higher quality? Bring your own key once, globally — **just paste the key, the provider is auto-detected from its prefix:**
 
 ```bash
-prjct embeddings set --key sk-...      # stored in the macOS Keychain (else a 0600 file), never in config
-prjct embeddings test                  # validate connectivity
-prjct embeddings status                # show provider + key location
+prjct embeddings set --key sk-or-v1-...   # → OpenRouter (auto-detected)
+prjct embeddings set --key sk-...         # → OpenAI (auto-detected)
+prjct embeddings test                     # validate connectivity (full error + hint on failure)
+prjct embeddings status                   # show provider, model, base URL + key location
+prjct embeddings clear                    # forget the key AND settings (falls back to local)
 ```
 
-One key applies to every project (OpenAI-compatible — OpenAI, Ollama at `http://localhost:11434/v1`, LM Studio, …). Without it, the local embedder is used. Each project re-vectorizes on its next session.
+One key applies to every project. **Any OpenAI-compatible `/embeddings` provider works.** For providers without a recognizable key prefix (or a custom/self-hosted endpoint), point `--base-url` at its root:
+
+```bash
+# Ollama (local, no key) / LM Studio / Together / Mistral / Voyage / Jina / DeepInfra …
+prjct embeddings set --model nomic-embed-text --base-url http://localhost:11434/v1
+```
+
+`set` is partial-update friendly — `set --key …` keeps your existing model and base URL; `set --model …` keeps your key. An explicit `--base-url` always overrides auto-detection.
+
+Providers that don't use `Authorization: Bearer` are supported too via auth flags — e.g. **Azure OpenAI** (`api-key` header + `api-version` query):
+
+```bash
+prjct embeddings set --key "$AZURE_KEY" \
+  --base-url https://RESOURCE.openai.azure.com/openai/deployments/DEPLOYMENT \
+  --auth-header api-key --auth-scheme none --query "api-version=2023-05-15"
+```
+
+| Flag | Purpose |
+| --- | --- |
+| `--key <k>` | API key — stored in the Keychain (else a 0600 file), never in config |
+| `--base-url <u>` | Provider's OpenAI-compatible root (default `https://api.openai.com/v1`) |
+| `--model <m>` | Model id (default `text-embedding-3-small`) |
+| `--auth-header <h>` | Header carrying the key (default `authorization`; `api-key` for Azure) |
+| `--auth-scheme <s\|none>` | Prefix before the key (default `Bearer`; `none` = raw key) |
+| `--headers "k=v,k2=v2"` | Extra static headers (gateways, attribution) |
+| `--query <qs>` | Raw query string appended to the URL (e.g. `api-version=2023-05-15`) |
+
+Without a key the built-in local embedder is used. Vector dimensionality is detected from the provider's response (no hardcoded size). Each project re-vectorizes on its next session.
 
 ### Drop files into the vault (bidirectional)