diff --git a/CHANGELOG.md b/CHANGELOG.md index 49de3e576..be5eed2ed 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -1,5 +1,20 @@ # Changelog +## [1.93.29.0] - 2026-11-03 + +**A `gbrain serve --http` on this machine's loopback address now counts as a local brain, so transcripts get imported again.** gbrain recommends this setup for local agents on a Postgres brain (`gbrain bootstrap harness`). gstack treated every http MCP entry as a brain on another machine (Path 4). So `/sync-gbrain` staged transcripts for a brain admin who doesn't exist and never ran `gbrain import` (#3539). Ported from Willy Morin's #3552. + +### Fixed +- **One locality rule** (`lib/gbrain-mcp-locality.ts`). An http, url or sse entry is local when its URL is http(s) on `localhost`, `127.0.0.0/8` or `::1`, and this machine's gbrain config names a Postgres engine (a `database_url`, not PGLite) with no `remote_mcp` marker. A loopback URL next to a local PGLite stays remote, so Path 4's split-engine setup through a tunnel is unchanged. A URL with userinfo or `${` is never local. Every other entry is classified as before. +- **`gstack-memory-ingest`** imports through gbrain in that case. Pages an earlier run left `staged` / `remote-http` count as new and are imported on the next local run. Pages older than the 90-day window need one `gstack-memory-ingest --bulk --all-history` run. +- **`gstack-gbrain-detect`** reports `local-stdio` for it in all three tiers, and `gbrain-local-status` and the `gstack-skill-start` banner use the same rule. A remote `(SSE)` line in `claude mcp list` no longer reads as local. With several gbrain servers, the exact `gbrain:` entry wins, as in the other tiers. + +### Docs +- `USING_GBRAIN_WITH_GSTACK.md`: a loopback serve over this machine's Postgres brain is not Path 4; a loopback tunnel next to a local PGLite still is. + +### Tests +- `test/gbrain-mcp-locality.test.ts` covers the rule, the list-line parser, and a check that the shipped jq copy never says local where the TypeScript rule says remote. The detect, local-status and memory-ingest tests add loopback and stuck-page cases. On main's code (the tests-first commit f3712618), 13 of the new cases fail and `test/gbrain-mcp-locality.test.ts` cannot load, because the module it tests is new. + ## [1.93.28.0] - 2026-11-03 **`test/cso-snapshot-state.test.ts` no longer hangs when a paused contender is resumed too early.** diff --git a/USING_GBRAIN_WITH_GSTACK.md b/USING_GBRAIN_WITH_GSTACK.md index 7b8d40cf4..3153388a0 100644 --- a/USING_GBRAIN_WITH_GSTACK.md +++ b/USING_GBRAIN_WITH_GSTACK.md @@ -74,6 +74,8 @@ If you accept the local PGLite, you end up in **split-engine mode**: - **Brain/context queries** (`mcp__gbrain__search`, `mcp__gbrain__query`, `mcp__gbrain__get_page`) route to the remote MCP. Plans, retros, learnings, cross-machine memory — all on the shared server. - **Code queries** (`gbrain code-def`, `code-refs`, `code-callers`, `code-callees`, `gbrain search` for code) route to the local PGLite via the `.gbrain-source` pin in each worktree. Indexed locally, fast, never leaves the machine. +A `gbrain serve --http` on this machine's loopback address (`localhost`, `127.0.0.1`, `::1`) over this machine's own Postgres brain is not Path 4. That is gbrain's recommended setup for local agents on a Postgres brain (`gbrain bootstrap harness`), and gstack treats it as a local brain: transcripts go through `gbrain import` here. gstack decides this from the URL and the local gbrain config alone, and only a config with a Postgres `database_url` counts. A loopback URL next to a local PGLite stays Path 4, so the split-engine setup above keeps working when you reach the remote brain through an SSH tunnel or port forward on a loopback port. + The two engines are independent. Wiping the local PGLite doesn't touch the remote brain; rotating the remote MCP bearer doesn't affect local code search. This is also the right configuration if your remote brain admin can't (or shouldn't) index every developer's checkout — local code stays local. ## MCP registration for Claude Code diff --git a/VERSION b/VERSION index 6e33e7843..636e6c73d 100644 --- a/VERSION +++ b/VERSION @@ -1 +1 @@ -1.93.28.0 +1.93.29.0 diff --git a/agents-digest/gstack-AGENTS.md b/agents-digest/gstack-AGENTS.md index b1b25db51..a5186e272 100644 --- a/agents-digest/gstack-AGENTS.md +++ b/agents-digest/gstack-AGENTS.md @@ -1,4 +1,4 @@ -# gstack digest v1.93.28.0 — regenerate/re-copy after upgrading gstack +# gstack digest v1.93.29.0 — regenerate/re-copy after upgrading gstack Behavioral rules from gstack (https://github.com/garrytan/gstack), compressed for agent hosts without a full skill install. The full skills add workflows, diff --git a/bin/gstack-gbrain-detect b/bin/gstack-gbrain-detect index e79f4d558..0183f549d 100755 --- a/bin/gstack-gbrain-detect +++ b/bin/gstack-gbrain-detect @@ -45,6 +45,7 @@ import { readGbrainVersion, } from "../lib/gbrain-local-status"; import { gbrainConfigDir, isTransactionModePooler } from "../lib/gbrain-exec"; +import { classifyGbrainMcpEntry, classifyGbrainMcpListLine, type GbrainMcpEntry } from "../lib/gbrain-mcp-locality"; import { resolveStateRoot } from "../lib/state-root"; const STATE_DIR = resolveStateRoot(); @@ -154,7 +155,19 @@ function detectBrainGit(): boolean { // 1. `claude mcp get gbrain --json` — public CLI surface, structured output // 2. `claude mcp list` text-grep — older claude versions without --json // 3. `~/.claude.json` jq read — last resort if `claude` isn't on PATH -function detectMcpMode(): "local-stdio" | "remote-http" | "none" { +// +// #3539: an HTTP entry on a loopback host whose local gbrain config has an +// engine and no remote_mcp marker is a resident `gbrain serve --http` of this +// machine's own brain. It reports "local-stdio" (the local-engine mode every +// consumer already understands: memory ingest imports, sync-gbrain skips the +// remote trust-policy question) rather than a new value. +type McpMode = "local-stdio" | "remote-http"; +function classify(entry: GbrainMcpEntry): McpMode | null { + const c = classifyGbrainMcpEntry(entry); + return c === "remote" ? "remote-http" : c === "local" ? "local-stdio" : null; +} + +function detectMcpMode(): McpMode | "none" { const claudeOnPath = tryExec("sh", ["-c", "command -v claude"], 1_000) !== null; if (claudeOnPath) { // Tier 1: `claude mcp get gbrain --json` @@ -167,11 +180,8 @@ function detectMcpMode(): "local-stdio" | "remote-http" | "none" { command?: string; url?: string; }; - const mtype = parsed.type || parsed.transport || ""; - if (mtype === "http" || mtype === "sse") return "remote-http"; - if (mtype === "stdio") return "local-stdio"; - if (parsed.url) return "remote-http"; - if (parsed.command) return "local-stdio"; + const c = classify(parsed); + if (c) return c; } catch { // fall through } @@ -179,34 +189,28 @@ function detectMcpMode(): "local-stdio" | "remote-http" | "none" { // Tier 2: `claude mcp list` text-grep. Name-pattern generalized (#2051): // a gbrain server registered as e.g. "gbrain-remote" or "gbrain_work" // still counts. Anchored to the gbrain token so unrelated servers can't - // false-positive. + // false-positive. The exact `gbrain:` line wins over a variant, the same + // order as tier 3 (and memory-ingest, which reads mcpServers.gbrain): + // {gbrain-a: loopback, gbrain: remote} is remote-http. const list = tryExec("claude", ["mcp", "list"], 3_000); if (list) { - const line = list.split("\n").find((l) => /^gbrain([-_][\w-]*)?:/.test(l)); + const lines = list.split("\n"); + const line = + lines.find((l) => l.startsWith("gbrain:")) ?? lines.find((l) => /^gbrain[-_][\w-]*:/.test(l)); if (line) { - if (/\b(http|HTTP)\b/.test(line)) return "remote-http"; - return "local-stdio"; + // The list line carries the URL for HTTP and SSE servers + // ("gbrain: http://127.0.0.1:3131/mcp (HTTP) - ✓ Connected"), so the + // loopback rule (#3539) applies here too. The whole URL token is + // classified, never cut at ")"; an HTTP or SSE line without a usable + // URL keeps the old remote-http answer. + return classifyGbrainMcpListLine(line) === "local" ? "local-stdio" : "remote-http"; } } } // Tier 3: read ~/.claude.json directly - interface McpServerEntry { - type?: string; - transport?: string; - command?: string; - url?: string; - } const cj = tryReadJSON(CLAUDE_JSON) as - | { mcpServers?: Record } + | { mcpServers?: Record } | null; - const classify = (entry: McpServerEntry): "local-stdio" | "remote-http" | null => { - const mtype = entry.type || entry.transport || ""; - if (mtype === "url" || mtype === "http" || mtype === "sse") return "remote-http"; - if (mtype === "stdio") return "local-stdio"; - if (entry.url) return "remote-http"; - if (entry.command) return "local-stdio"; - return null; - }; const servers = cj?.mcpServers || {}; const exact = servers["gbrain"]; if (exact) { diff --git a/bin/gstack-memory-ingest.ts b/bin/gstack-memory-ingest.ts index b2a69fe1a..dba4605fe 100644 --- a/bin/gstack-memory-ingest.ts +++ b/bin/gstack-memory-ingest.ts @@ -69,6 +69,7 @@ import { withErrorContext, } from "../lib/gstack-memory-helpers"; import { execGbrainText, gbrainConfigDir, spawnGbrain, spawnGbrainAsync } from "../lib/gbrain-exec"; +import { classifyGbrainMcpEntry, type GbrainMcpEntry } from "../lib/gbrain-mcp-locality"; import { constrainSourceId, parseSourcesList } from "../lib/gbrain-sources"; import { BISECT_AFTER_REFUSALS, @@ -427,9 +428,19 @@ function fileSha256(path: string): string { } } +/** + * #3539: a page staged for a remote brain's pull (`staged` / `remote-http`) + * was never imported here. Once the brain is local (a loopback serve, or the + * entry switched back to stdio), that page has not landed, so it must be + * prepared again instead of counting as unchanged. + */ +function stagedForRemotePull(entry: StateEntry | undefined): boolean { + return !!entry && entry.status === "staged" && entry.source_id === "remote-http" && !isRemoteHttpMcpMode(); +} + function fileChangedSinceState(path: string, state: IngestState, verifyHash = false): boolean { const entry = state.sessions[path]; - if (isRetryable(entry)) return true; + if (isRetryable(entry) || stagedForRemotePull(entry)) return true; try { const st = statSync(path); const mtimeNs = Math.floor(st.mtimeMs * 1e6); @@ -1418,7 +1429,7 @@ async function probeMode(args: CliArgs): Promise { totalBytes += size; const entry = state.sessions[path]; - if (!entry) newCount++; + if (!entry || stagedForRemotePull(entry)) newCount++; else if (fileChangedSinceState(path, state, args.scanSecrets)) updatedCount++; else unchangedCount++; } @@ -1800,23 +1811,27 @@ function makePersistentTranscriptDir(): string { * * Reads ~/.claude.json directly (same fallback chain as gstack-gbrain-detect * Tier 3). Cheap: one fs read, no fork-exec. + * + * #3539: an HTTP entry on a loopback host that reaches this machine's own + * gbrain engine (a resident `gbrain serve --http`) is not remote; see + * lib/gbrain-mcp-locality.ts. Memoized: the change check consults it per + * state entry, and ~/.claude.json can be megabytes. */ +let remoteHttpMcpMode: boolean | undefined; function isRemoteHttpMcpMode(): boolean { + remoteHttpMcpMode ??= readRemoteHttpMcpMode(); + return remoteHttpMcpMode; +} + +function readRemoteHttpMcpMode(): boolean { const home = process.env.HOME || homedir(); const claudeJsonPath = join(home, ".claude.json"); if (!existsSync(claudeJsonPath)) return false; try { const parsed = JSON.parse(readFileSync(claudeJsonPath, "utf-8")) as { - mcpServers?: { - gbrain?: { type?: string; transport?: string; url?: string }; - }; + mcpServers?: { gbrain?: GbrainMcpEntry }; }; - const entry = parsed.mcpServers?.gbrain; - if (!entry) return false; - const mtype = entry.type || entry.transport || ""; - if (mtype === "url" || mtype === "http" || mtype === "sse") return true; - if (entry.url) return true; - return false; + return classifyGbrainMcpEntry(parsed.mcpServers?.gbrain) === "remote"; } catch { return false; } diff --git a/bin/gstack-skill-start b/bin/gstack-skill-start index fb26ee8e5..16b1c9133 100755 --- a/bin/gstack-skill-start +++ b/bin/gstack-skill-start @@ -426,7 +426,29 @@ if command -v jq >/dev/null 2>&1 && [ -f "$HOME/.claude.json" ] && grep -q '"gbr _GBRAIN_MCP_ENTRY=$(jq -c --arg cwd "$PWD" '((.projects // {}) | to_entries | map(select((.key as $k | $cwd == $k or ($cwd | startswith($k + "/")) or ($cwd | startswith($k + "\\"))) and ((try .value.mcpServers.gbrain catch null) != null))) | sort_by(.key | length) | last | .value.mcpServers.gbrain) // .mcpServers.gbrain // empty' "$HOME/.claude.json" 2>/dev/null) _GBRAIN_MCP_TYPE=$(printf '%s' "$_GBRAIN_MCP_ENTRY" | jq -r '.type // .transport // empty' 2>/dev/null) case "$_GBRAIN_MCP_TYPE" in - url|http|sse) _GBRAIN_MCP_MODE="remote-http" ;; + url|http|sse) + _GBRAIN_MCP_MODE="remote-http" + # #3539: a loopback URL whose local gbrain config names a Postgres + # engine (a database_url, engine not pglite) and no remote_mcp marker is + # this machine's own `gbrain serve --http`, not a remote brain. A conservative subset of lib/gbrain-mcp-locality.ts, not + # an exact mirror: the URL test accepts only canonical loopback + # spellings (localhost, 127.a.b.c with valid decimal octets, [::1], a + # valid port), so other spellings stay remote here. It lowercases with + # ascii_downcase, not the "i" flag, which case-folds U+017F to "s". As in + # TS, a URL containing "${" is remote (Claude Code expands it), and the + # regex rejects userinfo. The config test + # matches the TS one: remote_mcp is a marker when JS-truthy, as in + # gbrain's isThinClient (null, false, 0 and "" are not), and + # database_url counts when non-blank after JS trim() and engine is not + # pglite. GBRAIN_HOME is the + # parent of .gbrain, as in gbrainConfigDir. test/gbrain-mcp-locality.test.ts + # runs both programs below against the TS rule. + if printf '%s' "$_GBRAIN_MCP_ENTRY" | jq -e '(.url // "") | ascii_downcase | (contains("${") | not) and test("\\Ahttps?://(localhost|127(\\.(25[0-5]|2[0-4][0-9]|1[0-9][0-9]|[1-9]?[0-9])){3}|\\[::1\\])(:([0-9]{0,4}|[1-5][0-9]{4}|6[0-4][0-9]{3}|65[0-4][0-9]{2}|655[0-2][0-9]|6553[0-5]))?([/?#]|\\z)")' >/dev/null 2>&1 \ + && jq -e '((.remote_mcp // false) | . == false or . == 0 or . == "") and .engine != "pglite" and ((.database_url // "") | (type == "string" and test("[^\\s\\x{feff}]|\\x{0085}")))' \ + "${GBRAIN_HOME:-$HOME}/.gbrain/config.json" >/dev/null 2>&1; then + _GBRAIN_MCP_MODE="local-stdio" + fi + ;; stdio) _GBRAIN_MCP_MODE="local-stdio" ;; esac fi diff --git a/lib/gbrain-local-status.ts b/lib/gbrain-local-status.ts index 7f18a8c46..1dbeb7caa 100644 --- a/lib/gbrain-local-status.ts +++ b/lib/gbrain-local-status.ts @@ -54,6 +54,7 @@ import { atomicWriteSync } from "./fs-atomic"; import { homedir } from "os"; import { dirname, join } from "path"; import { buildGbrainEnv, gbrainChildCwd, gbrainConfigDir, isExecTimeout, NEEDS_SHELL_ON_WINDOWS } from "./gbrain-exec"; +import { classifyGbrainMcpEntry } from "./gbrain-mcp-locality"; import { resolveStateRoot } from "./state-root"; export type LocalEngineStatus = @@ -187,16 +188,11 @@ export function hasRemoteOnlyGbrainMcp( } catch { return false; } - // Same classification rules as gstack-gbrain-detect's detectMcpMode tier 3, - // including the #2051 name generalization (gbrain, gbrain-remote, gbrain_work). - const classify = (entry: McpEntry): "remote" | "local" | null => { - const mtype = entry.type || entry.transport || ""; - if (mtype === "url" || mtype === "http" || mtype === "sse") return "remote"; - if (mtype === "stdio") return "local"; - if (entry.url) return "remote"; - if (entry.command) return "local"; - return null; - }; + // The classifier gstack-gbrain-detect's detectMcpMode uses, so a loopback + // `gbrain serve --http` over this machine's own engine is local here too + // (#3539): detect never prints local-stdio next to thin-client. Name + // generalization (#2051: gbrain, gbrain-remote, gbrain_work) is below. + const classify = (entry: McpEntry): "remote" | "local" | null => classifyGbrainMcpEntry(entry, env ?? process.env); /** Extract the gbrain-relevant entries from an mcpServers object. */ const gbrainEntries = (servers: unknown): Record => { const out: Record = {}; diff --git a/lib/gbrain-mcp-locality.ts b/lib/gbrain-mcp-locality.ts new file mode 100644 index 000000000..559f15fb9 --- /dev/null +++ b/lib/gbrain-mcp-locality.ts @@ -0,0 +1,139 @@ +/** + * gbrain-mcp-locality — is a registered gbrain MCP entry a remote brain? (#3539) + * + * An http/url/sse MCP entry used to mean "Path 4: the brain lives on another + * machine", so transcripts were staged for a brain admin's pull instead of + * going through `gbrain import`. gbrain's own recommended setup for local + * agents on a Postgres brain is one resident `gbrain serve --http` on + * loopback, registered by `gbrain bootstrap harness`. That entry is HTTP but + * the brain is on this machine, and no brain admin will ever index the + * staged pages. + * + * Rule: an HTTP entry is LOCAL when its URL is http(s) on a loopback host AND + * this machine's gbrain config names a Postgres engine (a `database_url`, + * engine not `pglite`) and no `remote_mcp` thin-client marker. Every other + * HTTP entry stays remote, as before. A PGLite config does not count: Path 4's + * split-engine setup pairs a remote brain with a local PGLite for code search, + * and a remote brain reached through a tunnel on a loopback port must stay + * remote there; a PGLite brain held open by a resident serve is engine-locked + * for imports anyway. stdio entries are local, as before. Any + * other URL scheme (ws:, file:) stays remote. The marker test is gbrain's own + * isThinClient() truthiness (`!!remote_mcp`), so `remote_mcp: 0` or `""` is + * not a marker. + * + * File reads only: no subprocess, no network (a classifier network probe is + * the #1964 pathology). + * + * bin/gstack-skill-start carries a jq copy of this rule for its hot path. It + * is a conservative subset, not an exact mirror: its URL regex accepts only + * canonical loopback spellings (localhost, 127.a.b.c without leading zeros, + * [::1]), so whenever jq says local this file says local too, while some + * spellings this file accepts (127.1, 0x7f.1, 127.0.0.1.) stay remote in jq. + * The config half matches exactly. test/gbrain-mcp-locality.test.ts runs both + * copies over the same table and enforces that relation. + */ + +import { readFileSync } from "fs"; +import { join } from "path"; +import { gbrainConfigDir } from "./gbrain-exec"; + +export interface GbrainMcpEntry { + type?: string; + transport?: string; + command?: string; + url?: string; +} + +/** + * True when the URL is http(s) and its host is localhost, 127.0.0.0/8 or ::1. + * A URL with userinfo, or containing `${`, is never loopback: Claude Code + * expands `${VAR}` in `url` before connecting, so `http://${X}@127.0.0.1/mcp` + * can dial any host. + */ +export function isLoopbackMcpUrl(url: string | undefined | null): boolean { + if (typeof url !== "string" || url.trim() === "" || url.includes("${")) return false; + let parsed: URL; + try { + parsed = new URL(url.trim()); + } catch { + return false; + } + if (parsed.protocol !== "http:" && parsed.protocol !== "https:") return false; + if (parsed.username || parsed.password) return false; + const host = parsed.hostname.toLowerCase(); + if (host === "localhost" || host === "::1" || host === "[::1]") return true; + // WHATWG URL normalizes IPv4 forms (127.1, 0x7f.1) to dotted quads. + return /^127\.\d{1,3}\.\d{1,3}\.\d{1,3}$/.test(host); +} + +/** + * This machine's gbrain config (gbrainConfigDir()/config.json, honoring + * GBRAIN_HOME) names a Postgres engine (a non-blank database_url, engine not + * pglite) and carries no remote_mcp marker (truthy, as in gbrain's + * isThinClient). + */ +export function hasLocalPostgresEngine(env: NodeJS.ProcessEnv = process.env): boolean { + try { + const cfg = JSON.parse(readFileSync(join(gbrainConfigDir(env), "config.json"), "utf-8")) as { + engine?: unknown; + database_url?: unknown; + remote_mcp?: unknown; + } | null; + if (!cfg || typeof cfg !== "object") return false; + if (cfg.remote_mcp) return false; + return cfg.engine !== "pglite" && typeof cfg.database_url === "string" && cfg.database_url.trim() !== ""; + } catch { + return false; + } +} + +/** An HTTP MCP URL that reaches this machine's own Postgres gbrain engine. */ +export function isLocalHttpGbrain(url: string | undefined | null, env: NodeJS.ProcessEnv = process.env): boolean { + return isLoopbackMcpUrl(url) && hasLocalPostgresEngine(env); +} + +/** + * Classify one MCP registration: "remote" for an HTTP brain on another + * machine, "local" for stdio or a loopback HTTP serve of the local engine, + * null when the entry carries no transport signal. + */ +export function classifyGbrainMcpEntry( + entry: GbrainMcpEntry | null | undefined, + env: NodeJS.ProcessEnv = process.env, +): "local" | "remote" | null { + if (!entry || typeof entry !== "object") return null; + // Same precedence as the transport-only rule this replaces: explicit HTTP + // type, explicit stdio, then a bare url, then a bare command. + const mtype = entry.type || entry.transport || ""; + const remoteUnlessLoopback = (): "local" | "remote" => (isLocalHttpGbrain(entry.url, env) ? "local" : "remote"); + if (mtype === "url" || mtype === "http" || mtype === "sse") return remoteUnlessLoopback(); + if (mtype === "stdio") return "local"; + if (entry.url) return remoteUnlessLoopback(); + if (entry.command) return "local"; + return null; +} + +/** + * Classify the gbrain line of `claude mcp list`. HTTP and SSE servers print + * as `gbrain: (HTTP) - ✓ Connected` (or `(SSE)`); stdio servers print + * their command with no transport marker. + * + * The URL token is everything between the name and the LAST ` (HTTP)` / + * ` (SSE)` marker. It is never cut at `)`: a cut turns + * `http://127.0.0.1)@evil.com/mcp` (host evil.com) into `http://127.0.0.1`. + * A token containing whitespace, or one that is not a loopback http(s) URL + * over a local engine, is remote. A line that names HTTP/SSE but has no + * such token is remote, as before; a line with neither is stdio (local). + */ +export function classifyGbrainMcpListLine( + line: string, + env: NodeJS.ProcessEnv = process.env, +): "local" | "remote" { + const m = /^gbrain(?:[-_][\w-]*)?:\s+(.+)\s+\((?:HTTP|SSE)\)/.exec(line); + if (m) { + const token = m[1]; + return !/\s/.test(token) && isLocalHttpGbrain(token, env) ? "local" : "remote"; + } + if (/\b(?:http|HTTP)\b/.test(line) || /\((?:SSE|sse)\)/.test(line)) return "remote"; + return "local"; +} diff --git a/package.json b/package.json index 402ae4f9d..54ba56c66 100644 --- a/package.json +++ b/package.json @@ -1,6 +1,6 @@ { "name": "gstack", - "version": "1.93.28", + "version": "1.93.29", "description": "Garry's Stack — Claude Code skills + fast headless browser. One repo, one install, entire AI engineering workflow.", "license": "MIT", "type": "module", diff --git a/test/gbrain-local-status.test.ts b/test/gbrain-local-status.test.ts index 9d666a35e..038f66203 100644 --- a/test/gbrain-local-status.test.ts +++ b/test/gbrain-local-status.test.ts @@ -760,6 +760,50 @@ describe("lib/gbrain-local-status — bearer-token thin-client (#2520)", () => { expect(localEngineStatus({ noCache: true })).toBe("engine-locked"); }); + // #3539: a loopback `gbrain serve --http` over this machine's own engine is + // a local registration. Before, the transport-only copy of the rule called + // it remote, so detect printed gbrain_mcp_mode local-stdio next to + // gbrain_local_status thin-client for the same entry. + it("keeps the local engine's own status for a loopback http gbrain MCP over a local Postgres (#3539)", () => { + env = makeEnv({ + withGbrain: true, + gbrainBehavior: "engine-locked-v43", + withConfig: true, + claudeJson: { mcpServers: { gbrain: { type: "http", url: "http://127.0.0.1:3131/mcp" } } }, + }); + restoreEnv = applyEnv(env); + writeFileSync(env.configPath, JSON.stringify({ engine: "postgres", database_url: "postgres://fake" })); + expect(localEngineStatus({ noCache: true })).toBe("broken-db"); + }); + + it("still reclassifies a locked local PGLite as 'thin-client' behind a loopback http entry (Path 4 split-engine, #3539)", () => { + env = makeEnv({ + withGbrain: true, + gbrainBehavior: "engine-locked-v43", + withConfig: true, + claudeJson: { mcpServers: { gbrain: { type: "http", url: "http://127.0.0.1:3131/mcp" } } }, + }); + restoreEnv = applyEnv(env); + expect(localEngineStatus({ noCache: true })).toBe("thin-client"); + }); + + it("still reclassifies as 'thin-client' when the loopback entry's config carries remote_mcp (#3539 guard)", () => { + env = makeEnv({ + withGbrain: true, + gbrainBehavior: "engine-locked", + withConfig: true, + claudeJson: { mcpServers: { gbrain: { type: "http", url: "http://127.0.0.1:3131/mcp" } } }, + }); + restoreEnv = applyEnv(env); + writeFileSync( + env.configPath, + JSON.stringify({ engine: "pglite", remote_mcp: { mcp_url: "http://127.0.0.1:3131/mcp" } }), + ); + // Config marker short-circuits to thin-client before the probe; the + // loopback entry must not turn that into a local registration. + expect(localEngineStatus({ noCache: true })).toBe("thin-client"); + }); + it("still returns 'missing-config' when no gbrain MCP registration exists (discriminator)", () => { env = makeEnv({ withGbrain: true, diff --git a/test/gbrain-mcp-locality.test.ts b/test/gbrain-mcp-locality.test.ts new file mode 100644 index 000000000..3e81727e7 --- /dev/null +++ b/test/gbrain-mcp-locality.test.ts @@ -0,0 +1,378 @@ +/** + * #3539: a loopback `gbrain serve --http` entry over this machine's own gbrain + * engine is a local brain, not a Path 4 remote one. + * + * Pins lib/gbrain-mcp-locality.ts, and checks the jq copy in + * bin/gstack-skill-start (the preamble's hot path reads ~/.claude.json with + * jq, not through bun) against it: a few end-to-end skill-start cases, plus a + * table-driven parity run of the two jq programs themselves. + */ +import { describe, test, expect, beforeEach, afterEach } from 'bun:test'; +import { execFileSync } from 'child_process'; +import * as fs from 'fs'; +import * as os from 'os'; +import * as path from 'path'; +import { + classifyGbrainMcpEntry, + classifyGbrainMcpListLine, + hasLocalPostgresEngine, + isLoopbackMcpUrl, +} from '../lib/gbrain-mcp-locality'; + +const ROOT = path.resolve(import.meta.dir, '..'); +const START = path.join(ROOT, 'bin', 'gstack-skill-start'); + +let home: string; + +const PG = { engine: 'postgres', database_url: 'postgresql://u@localhost:5432/brain' }; + +function writeGbrainConfig(cfg: unknown, dir = home): void { + fs.mkdirSync(path.join(dir, '.gbrain'), { recursive: true }); + fs.writeFileSync(path.join(dir, '.gbrain', 'config.json'), JSON.stringify(cfg)); +} + +beforeEach(() => { + home = fs.mkdtempSync(path.join(os.tmpdir(), 'gbrain-locality-')); +}); + +afterEach(() => { + fs.rmSync(home, { recursive: true, force: true }); +}); + +describe('isLoopbackMcpUrl', () => { + test.each([ + 'http://127.0.0.1:3131/mcp', + 'http://127.8.9.10/mcp', + 'http://localhost:3131/mcp', + 'HTTP://LOCALHOST/mcp', + 'http://[::1]:3131/mcp', + ])('%s is loopback', (url) => { + expect(isLoopbackMcpUrl(url)).toBe(true); + }); + + test.each([ + 'https://brain.example.com/mcp', + 'http://127.0.0.1.evil.example/mcp', + 'http://localhost.evil.example/mcp', + 'http://10.0.0.5:3131/mcp', + 'http://0.0.0.0:3131/mcp', + 'ws://127.0.0.1:3131/mcp', + 'wss://localhost/mcp', + 'file://localhost/etc/passwd', + 'not a url', + '', + // Userinfo and ${VAR}: Claude Code expands ${VAR} in the url before it + // connects, so the host it dials is not the one written here. + 'http://evil@127.0.0.1/mcp', + 'http://user:PASSWORD@localhost:3131/mcp', + 'http://:pw@127.0.0.1/mcp', + 'http://${GBRAIN_X}@127.0.0.1:3131/mcp', + 'http://127.0.0.1:3131/${GBRAIN_X}', + ])('%s is not loopback', (url) => { + expect(isLoopbackMcpUrl(url)).toBe(false); + }); +}); + +describe('hasLocalPostgresEngine', () => { + test('a database_url on a non-PGLite engine, with no remote_mcp marker', () => { + const env = { HOME: home }; + expect(hasLocalPostgresEngine(env)).toBe(false); // no config + writeGbrainConfig(PG); + expect(hasLocalPostgresEngine(env)).toBe(true); + writeGbrainConfig({ database_url: 'postgresql://db.example.com/brain' }); + expect(hasLocalPostgresEngine(env)).toBe(true); + // PGLite (Path 4's split-engine code-search brain) never counts. + writeGbrainConfig({ engine: 'pglite', database_path: '/x/brain.pglite' }); + expect(hasLocalPostgresEngine(env)).toBe(false); + writeGbrainConfig({ engine: 'pglite', database_url: 'postgresql://u@localhost:5432/brain' }); + expect(hasLocalPostgresEngine(env)).toBe(false); + writeGbrainConfig({ ...PG, remote_mcp: { mcp_url: 'http://127.0.0.1:3131/mcp' } }); + expect(hasLocalPostgresEngine(env)).toBe(false); + writeGbrainConfig({ database_url: ' ' }); + expect(hasLocalPostgresEngine(env)).toBe(false); + // gbrain's isThinClient is `!!remote_mcp`: falsy values are not markers. + for (const remote_mcp of [null, false, 0, '']) { + writeGbrainConfig({ ...PG, remote_mcp }); + expect(hasLocalPostgresEngine(env)).toBe(true); + } + for (const remote_mcp of [{}, [], 'x', 1, true]) { + writeGbrainConfig({ ...PG, remote_mcp }); + expect(hasLocalPostgresEngine(env)).toBe(false); + } + }); + + test('honors GBRAIN_HOME as the parent of .gbrain, like gbrainConfigDir', () => { + const gbrainHome = path.join(home, 'elsewhere'); + writeGbrainConfig(PG, gbrainHome); + expect(hasLocalPostgresEngine({ HOME: home })).toBe(false); + expect(hasLocalPostgresEngine({ HOME: home, GBRAIN_HOME: gbrainHome })).toBe(true); + }); +}); + +describe('classifyGbrainMcpEntry', () => { + const loopback = { type: 'http', url: 'http://127.0.0.1:3131/mcp' }; + + test('loopback http entry over a local Postgres engine is local', () => { + writeGbrainConfig(PG); + expect(classifyGbrainMcpEntry(loopback, { HOME: home })).toBe('local'); + expect(classifyGbrainMcpEntry({ url: 'http://localhost:3131/mcp' }, { HOME: home })).toBe('local'); + }); + + test('loopback http entry stays remote without a local engine or with remote_mcp', () => { + expect(classifyGbrainMcpEntry(loopback, { HOME: home })).toBe('remote'); + writeGbrainConfig({ ...PG, remote_mcp: { mcp_url: 'http://127.0.0.1:3131/mcp' } }); + expect(classifyGbrainMcpEntry(loopback, { HOME: home })).toBe('remote'); + }); + + test('loopback http entry over a local PGLite stays remote (Path 4 split-engine through a tunnel)', () => { + writeGbrainConfig({ engine: 'pglite', database_path: '/x/brain.pglite' }); + expect(classifyGbrainMcpEntry(loopback, { HOME: home })).toBe('remote'); + }); + + test('non-loopback http entry stays remote even with a local engine', () => { + writeGbrainConfig(PG); + for (const type of ['http', 'sse', 'url']) { + expect(classifyGbrainMcpEntry({ type, url: 'https://brain.example.com/mcp' }, { HOME: home })).toBe('remote'); + } + }); + + test('stdio and command entries are local; empty entries carry no signal', () => { + expect(classifyGbrainMcpEntry({ type: 'stdio', command: 'gbrain' }, { HOME: home })).toBe('local'); + expect(classifyGbrainMcpEntry({ command: 'gbrain' }, { HOME: home })).toBe('local'); + expect(classifyGbrainMcpEntry({}, { HOME: home })).toBeNull(); + expect(classifyGbrainMcpEntry(undefined, { HOME: home })).toBeNull(); + }); +}); + +describe('classifyGbrainMcpListLine (`claude mcp list`, detect tier 2)', () => { + beforeEach(() => writeGbrainConfig(PG)); + + const LINES: Array<[string, 'local' | 'remote']> = [ + ['gbrain: http://127.0.0.1:3131/mcp (HTTP) - ✓ Connected', 'local'], + ['gbrain: http://localhost:3131/sse (SSE) - ✓ Connected', 'local'], + ['gbrain-local: http://[::1]:3131/mcp (HTTP) - ✗ Failed to connect', 'local'], + ['gbrain: http://127.0.0.1:3131/mcp(x) (HTTP) - ✓ Connected', 'local'], + ['gbrain: /usr/local/bin/gbrain serve - ✓ Connected', 'local'], + // A URL cut at ")" or at whitespace would read these as loopback. + ['gbrain: http://127.0.0.1)@evil.com/mcp (HTTP) - ✓ Connected', 'remote'], + ['gbrain: http://localhost)@evil.com/mcp (HTTP) - ✓ Connected', 'remote'], + ['gbrain: http://127.0.0.1 (HTTP)@evil.com/mcp (HTTP) - ✓ Connected', 'remote'], + ['gbrain: http://127.0.0.1\t.evil.com/mcp (HTTP) - ✓ Connected', 'remote'], + ['gbrain: http://127.0.0.1 .evil.com/mcp (HTTP) - ✓ Connected', 'remote'], + ['gbrain: ws://127.0.0.1:3131/mcp (HTTP) - ✓ Connected', 'remote'], + ['gbrain: https://brain.example.com/mcp (HTTP) - ✓ Connected', 'remote'], + ['gbrain: https://brain.example.com/sse (SSE) - ✓ Connected', 'remote'], + ['gbrain: (SSE)', 'remote'], + ['gbrain: HTTP server - ✓ Connected', 'remote'], + // Claude Code flattens a newline/CR (and U+2028/2029, NEL, VT, FF) inside + // a url to a space when it prints the list, so an injected "second line" + // stays inside this one line and its URL token contains whitespace. + // These are the lines Claude Code 2.1.311 printed for such urls. + ['gbrain: http://127.0.0.1 (HTTP) @evil.invalid/mcp (HTTP) - ✘ Failed to connect', 'remote'], + ['gbrain: https://evil.invalid/a gbrain: http://127.0.0.1:1/mcp (HTTP) - ✓ Connected (HTTP) - ✘ Failed to connect', 'remote'], + ]; + + test.each(LINES)('%j → %s', (line: string, want: 'local' | 'remote') => { + expect(classifyGbrainMcpListLine(line, { HOME: home })).toBe(want); + }); + + test('a loopback list line without a local engine is remote', () => { + fs.rmSync(path.join(home, '.gbrain'), { recursive: true, force: true }); + expect(classifyGbrainMcpListLine('gbrain: http://127.0.0.1:3131/mcp (HTTP) - ✓ Connected', { HOME: home })).toBe('remote'); + }); +}); + +// Table-driven parity: the two jq programs bin/gstack-skill-start runs (read +// out of the script, so this checks the shipped text) against the TS rule. +// The config half must match exactly. The URL half is a conservative subset: +// jq may say remote where TS says local (non-canonical loopback spellings such +// as 127.1, 0x7f.1 or 127.0.0.1. that jq's regex does not accept), and +// must never say local where TS says remote. Exact URL parity would mean +// reimplementing WHATWG host parsing in jq; those spellings fall back to the +// old remote-http answer, which is the safe side. +describe('gstack-skill-start jq copy vs the TS rule (table-driven parity)', () => { + const src = fs.readFileSync(START, 'utf-8'); + const programs = [...src.matchAll(/jq -e '([^']+)'/g)].map((m) => m[1]); + const JQ_URL = programs.find((p) => p.startsWith('(.url // "")')); + const JQ_CFG = programs.find((p) => p.includes('remote_mcp')); + + function jqTrue(program: string, input: string): boolean { + try { + execFileSync('jq', ['-e', program], { input, stdio: ['pipe', 'ignore', 'ignore'], timeout: 30_000 }); + return true; + } catch { + return false; + } + } + + test('both jq programs are found in bin/gstack-skill-start', () => { + expect(JQ_URL).toBeDefined(); + expect(JQ_CFG).toBeDefined(); + }); + + // [url, expected TS answer]. jq must never be true where TS is false. + const URLS: Array<[string, boolean]> = [ + ['http://127.0.0.1:3131/mcp', true], + ['http://localhost:3131/mcp', true], + ['HTTP://LOCALHOST/mcp', true], + ['https://127.0.0.1/mcp', true], + ['http://[::1]:3131/mcp', true], + ['http://127.255.255.255:1/x', true], + ['http://127.0.0.1', true], + ['http://127.0.0.1:3131?x', true], + ['http://localhost:65535/', true], + ['http://127.0.0.1/mcp ', true], + // Non-canonical loopback spellings: TS local, jq remote (safe direction). + ['http://127.1/mcp', true], + ['http://0x7f.1/mcp', true], + ['http://2130706433/mcp', true], + ['http://[0:0:0:0:0:0:0:1]/mcp', true], + ['http://127.0.0.1.:3131/mcp', true], + [' http://127.0.0.1/mcp', true], + // WHATWG IDNA maps U+017F to "s": TS local; jq's ascii_downcase does not. + ['http://localhoſt/mcp', true], + // Userinfo and ${VAR} (expanded by Claude Code before it connects): remote. + ['http://evil@127.0.0.1/mcp', false], + ['http://user:PASSWORD@localhost:3131/mcp', false], + ['http://${GBRAIN_X}@127.0.0.1:3131/mcp', false], + ['http://127.0.0.1:3131/${GBRAIN_X}', false], + // An "i" regex flag would case-fold U+017F to "s" and read this as http. + ['httpſ://127.0.0.1/mcp', false], + ['HTTPſ://LOCALHOST/mcp', false], + // Remote hosts: both must say remote. + ['http://127.0.0.1)@evil.com/mcp', false], + ['http://localhost)@evil.com/mcp', false], + ['http://127.0.0.1@evil.com/mcp', false], + ['http://localhost:3131' + '@evil.com/mcp', false], + ['http://127.0.0.1:3131' + '@evil.com/mcp', false], + ['http://127.0.0.1.evil.example/mcp', false], + ['http://localhost.evil.example/mcp', false], + ['http://127.0.0.1.nip.io/mcp', false], + ['http://127.0.0.1%2eevil.com/', false], + ['http://127.0.0.1\tevil.com/', false], + ['http://127.0.0.1\nevil.com/', false], + ['http://0.0.0.0:3131/mcp', false], + ['http://[::ffff:127.0.0.1]/mcp', false], + // Invalid hosts/ports: cannot connect; jq used to say local here. + ['http://127.0.0.999/mcp', false], + ['http://127.0.0.256/mcp', false], + ['http://127.0.0.1:99999/mcp', false], + ['http://localhost:65536/mcp', false], + // Non-http schemes. + ['ws://127.0.0.1/mcp', false], + ['wss://localhost/mcp', false], + ['file://localhost/etc', false], + ['https://brain.example.com/mcp', false], + ['not a url', false], + ['', false], + ]; + + test.each(URLS)('url %j: TS %p, jq never local where TS is remote', (url: string, tsWant: boolean) => { + const ts = isLoopbackMcpUrl(url); + expect(ts).toBe(tsWant); + if (!ts) expect(jqTrue(JQ_URL!, JSON.stringify({ url }))).toBe(false); + }); + + test('jq agrees with TS on the canonical spellings gbrain bootstrap writes', () => { + for (const url of ['http://127.0.0.1:3131/mcp', 'http://localhost:3131/mcp', 'http://[::1]:3131/mcp']) { + expect(jqTrue(JQ_URL!, JSON.stringify({ url }))).toBe(true); + } + }); + + const CONFIGS: unknown[] = [ + { engine: 'pglite' }, + { engine: 'pglite', database_url: 'postgresql://u@localhost:5432/brain' }, + { database_url: 'postgresql://u@localhost:5432/brain' }, + { engine: 'postgres', database_url: 'postgresql://u@localhost:5432/brain', remote_mcp: false }, + { engine: 'postgres', database_url: 'postgresql://db.example.com/x' }, + { database_url: ' ' }, + { database_url: '\n\t' }, + { database_url: '\u00a0\u2028\u3000\ufeff' }, + { database_url: '\u0085' }, + { database_url: 5 }, + { engine: 'pglite', remote_mcp: { mcp_url: 'http://127.0.0.1:3131/mcp' } }, + { engine: 'pglite', remote_mcp: {} }, + { engine: 'pglite', remote_mcp: [] }, + { engine: 'pglite', remote_mcp: 'x' }, + { engine: 'pglite', remote_mcp: 1 }, + { engine: 'pglite', remote_mcp: true }, + { engine: 'pglite', remote_mcp: false }, + { engine: 'pglite', remote_mcp: 0 }, + { engine: 'pglite', remote_mcp: '' }, + { engine: 'pglite', remote_mcp: null }, + { engine: 'PGLite' }, + {}, + null, + [], + 'str', + ]; + + test.each(CONFIGS.map((c) => [c]))('config %j: jq matches TS exactly', (cfg: unknown) => { + writeGbrainConfig(cfg); + const ts = hasLocalPostgresEngine({ HOME: home }); + const jq = jqTrue(JQ_CFG!, fs.readFileSync(path.join(home, '.gbrain', 'config.json'), 'utf-8')); + expect(jq).toBe(ts); + }); + + test('malformed config: both say no local engine', () => { + fs.mkdirSync(path.join(home, '.gbrain'), { recursive: true }); + fs.writeFileSync(path.join(home, '.gbrain', 'config.json'), '{not json'); + expect(hasLocalPostgresEngine({ HOME: home })).toBe(false); + expect(jqTrue(JQ_CFG!, '{not json')).toBe(false); + }); +}); + +// The skill-start preamble prints `ARTIFACTS_SYNC: remote-mode (...)` only +// when it classifies the gbrain entry as remote-http; otherwise a cold home +// prints `ARTIFACTS_SYNC: off`. +describe('gstack-skill-start mirrors the rule', () => { + function artifactsSyncLine(entry: unknown, gbrainConfig?: unknown, extraEnv: Record = {}): string { + const gstackHome = path.join(home, '.gstack'); + fs.mkdirSync(gstackHome, { recursive: true }); + fs.writeFileSync(path.join(gstackHome, 'config.yaml'), 'update_check: false\n'); + fs.writeFileSync(path.join(home, '.claude.json'), JSON.stringify({ mcpServers: { gbrain: entry } })); + if (gbrainConfig !== undefined) writeGbrainConfig(gbrainConfig, extraEnv.GBRAIN_HOME ?? home); + const out = execFileSync(START, ['--skill', 'testskill'], { + timeout: 30_000, + encoding: 'utf-8', + cwd: home, + env: { + PATH: process.env.PATH!, + HOME: home, + GSTACK_HOME: gstackHome, + GSTACK_LEASE_DIR: path.join(gstackHome, 'leases'), + ...extraEnv, + }, + }); + return out.split('\n').find((l) => l.startsWith('ARTIFACTS_SYNC:')) ?? ''; + } + + test('loopback http entry over a local Postgres engine is not remote-mode', () => { + expect(artifactsSyncLine({ type: 'http', url: 'http://127.0.0.1:3131/mcp' }, PG)).toBe('ARTIFACTS_SYNC: off'); + }); + + test('loopback http entry over a local PGLite stays remote-mode (Path 4 split-engine)', () => { + expect(artifactsSyncLine({ type: 'http', url: 'http://127.0.0.1:3131/mcp' }, { engine: 'pglite' })).toMatch( + /^ARTIFACTS_SYNC: remote-mode \(managed by brain server 127\.0\.0\.1\)$/, + ); + }); + + test('GBRAIN_HOME locates the local engine config', () => { + const gbrainHome = path.join(home, 'gb'); + expect( + artifactsSyncLine({ type: 'http', url: 'http://localhost:3131/mcp' }, PG, { GBRAIN_HOME: gbrainHome }), + ).toBe('ARTIFACTS_SYNC: off'); + }); + + test('loopback http entry with remote_mcp in the config stays remote-mode', () => { + expect( + artifactsSyncLine({ type: 'http', url: 'http://127.0.0.1:3131/mcp' }, { ...PG, remote_mcp: { mcp_url: 'x' } }), + ).toMatch(/^ARTIFACTS_SYNC: remote-mode \(managed by brain server 127\.0\.0\.1\)$/); + }); + + test('non-loopback http entry stays remote-mode', () => { + expect(artifactsSyncLine({ type: 'http', url: 'https://brain.example.com/mcp' }, PG)).toMatch( + /^ARTIFACTS_SYNC: remote-mode \(managed by brain server brain\.example\.com\)$/, + ); + }); +}); diff --git a/test/gstack-gbrain-detect-mcp-mode.test.ts b/test/gstack-gbrain-detect-mcp-mode.test.ts index e1cfba204..e1ebf6c3d 100644 --- a/test/gstack-gbrain-detect-mcp-mode.test.ts +++ b/test/gstack-gbrain-detect-mcp-mode.test.ts @@ -266,6 +266,168 @@ describe('gbrain_mcp_mode — Tier 3: ~/.claude.json jq read', () => { }); }); +// #3539: a loopback `gbrain serve --http` over this machine's own engine is a +// local brain. It reports local-stdio (the local-engine mode every consumer +// already understands), in all three tiers. +describe('gbrain_mcp_mode — loopback serve --http over a local engine (#3539)', () => { + const LOOPBACK = 'http://127.0.0.1:3131/mcp'; + function localEngine(extra: Record = {}, dir = tmpHome): void { + fs.mkdirSync(path.join(dir, '.gbrain'), { recursive: true }); + fs.writeFileSync( + path.join(dir, '.gbrain', 'config.json'), + JSON.stringify({ engine: 'postgres', database_url: 'postgresql://u@localhost:5432/brain', ...extra }), + ); + } + + test('tier 1: loopback http entry + local engine → local-stdio', () => { + localEngine(); + makeFakeClaude({ hasGetJson: true, getJsonOutput: JSON.stringify({ type: 'http', url: LOOPBACK }) }); + expect(runDetect().json.gbrain_mcp_mode).toBe('local-stdio'); + }); + + test('tier 2: loopback URL on the mcp list line + local engine → local-stdio', () => { + localEngine(); + makeFakeClaude({ hasGetJson: false, hasMcpList: true, mcpListOutput: `gbrain: ${LOOPBACK} (HTTP) - ✓ Connected` }); + expect(runDetect().json.gbrain_mcp_mode).toBe('local-stdio'); + }); + + test('tier 3: loopback http entry + local engine under GBRAIN_HOME → local-stdio', () => { + const gbrainHome = path.join(tmpHome, 'gb'); + localEngine({}, gbrainHome); + fs.writeFileSync( + path.join(tmpHome, '.claude.json'), + JSON.stringify({ mcpServers: { gbrain: { type: 'http', url: LOOPBACK, headers: { Authorization: 'Bearer x' } } } }) + ); + expect(runDetect({ GBRAIN_HOME: gbrainHome }).json.gbrain_mcp_mode).toBe('local-stdio'); + }); + + test('loopback entry without a local engine config stays remote-http', () => { + fs.writeFileSync(path.join(tmpHome, '.claude.json'), JSON.stringify({ mcpServers: { gbrain: { type: 'http', url: LOOPBACK } } })); + expect(runDetect().json.gbrain_mcp_mode).toBe('remote-http'); + }); + + test('loopback entry with remote_mcp in the config stays remote-http', () => { + localEngine({ remote_mcp: { mcp_url: 'https://team-brain.example.com/mcp' } }); + makeFakeClaude({ hasGetJson: false, hasMcpList: true, mcpListOutput: `gbrain: ${LOOPBACK} (HTTP) - ✓ Connected` }); + expect(runDetect().json.gbrain_mcp_mode).toBe('remote-http'); + }); + + test('non-loopback entry with a local engine stays remote-http', () => { + localEngine(); + makeFakeClaude({ hasGetJson: true, getJsonOutput: JSON.stringify({ type: 'http', url: 'https://brain.example.com/mcp' }) }); + expect(runDetect().json.gbrain_mcp_mode).toBe('remote-http'); + }); + + // Tier 2 classifies the whole URL token. A URL cut at ")" or at whitespace + // would read these remote hosts as 127.0.0.1 / localhost. + test.each([ + 'gbrain: http://127.0.0.1)@evil.com/mcp (HTTP) - ✓ Connected', + 'gbrain: http://localhost)@evil.com/mcp (HTTP) - ✓ Connected', + 'gbrain: http://127.0.0.1 (HTTP)@evil.com/mcp (HTTP) - ✓ Connected', + 'gbrain: http://127.0.0.1\t.evil.com/mcp (HTTP) - ✓ Connected', + 'gbrain: ws://127.0.0.1:3131/mcp (HTTP) - ✓ Connected', + 'gbrain: http://127.0.0.1.evil.example/mcp (HTTP) - ✓ Connected', + 'gbrain: HTTP server, url hidden - ✓ Connected', + ])('tier 2: %s → remote-http (local engine present)', (line: string) => { + localEngine(); + makeFakeClaude({ hasGetJson: false, hasMcpList: true, mcpListOutput: line }); + expect(runDetect().json.gbrain_mcp_mode).toBe('remote-http'); + }); + + test('tier 2: a loopback URL with a ")" in its path is still local', () => { + localEngine(); + makeFakeClaude({ hasGetJson: false, hasMcpList: true, mcpListOutput: 'gbrain: http://127.0.0.1:3131/mcp(x) (HTTP) - ✓ Connected' }); + expect(runDetect().json.gbrain_mcp_mode).toBe('local-stdio'); + }); + + // (SSE) lines carry no "http" word. Before, they fell through to + // local-stdio even for a remote host. + test('tier 2: remote (SSE) line → remote-http', () => { + localEngine(); + makeFakeClaude({ hasGetJson: false, hasMcpList: true, mcpListOutput: 'gbrain: https://brain.example.com/sse (SSE) - ✓ Connected' }); + expect(runDetect().json.gbrain_mcp_mode).toBe('remote-http'); + }); + + test('tier 2: loopback (SSE) line over a local engine → local-stdio', () => { + localEngine(); + makeFakeClaude({ hasGetJson: false, hasMcpList: true, mcpListOutput: 'gbrain: http://localhost:3131/sse (SSE) - ✓ Connected' }); + expect(runDetect().json.gbrain_mcp_mode).toBe('local-stdio'); + }); + + // Several gbrain servers: the exact `gbrain:` line wins whatever the list + // order (the .claude.json insertion order), as in tier 3 and memory-ingest. + test.each([ + ['variant first', `gbrain-a: ${LOOPBACK} (HTTP) - ✓ Connected\ngbrain: https://team.example.com/mcp (HTTP) - ✓ Connected`], + ['exact first', `gbrain: https://team.example.com/mcp (HTTP) - ✓ Connected\ngbrain-a: ${LOOPBACK} (HTTP) - ✓ Connected`], + ])('tier 2: loopback gbrain-a + remote gbrain (%s) → remote-http', (_order: string, list: string) => { + localEngine(); + makeFakeClaude({ hasGetJson: false, hasMcpList: true, mcpListOutput: list }); + expect(runDetect().json.gbrain_mcp_mode).toBe('remote-http'); + }); + + test.each([ + ['variant first', `gbrain_work: https://team.example.com/mcp (HTTP) - ✓ Connected\ngbrain: ${LOOPBACK} (HTTP) - ✓ Connected`], + ['exact first', `gbrain: ${LOOPBACK} (HTTP) - ✓ Connected\ngbrain_work: https://team.example.com/mcp (HTTP) - ✓ Connected`], + ])('tier 2: loopback gbrain + remote gbrain_work (%s) → local-stdio', (_order: string, list: string) => { + localEngine(); + makeFakeClaude({ hasGetJson: false, hasMcpList: true, mcpListOutput: list }); + expect(runDetect().json.gbrain_mcp_mode).toBe('local-stdio'); + }); + + test('tier 2: with no exact gbrain line, the first gbrain[-_]* line counts', () => { + localEngine(); + makeFakeClaude({ + hasGetJson: false, + hasMcpList: true, + mcpListOutput: `posthog: https://mcp.posthog.com/mcp (HTTP)\ngbrain-remote: https://team.example.com/mcp (HTTP) - ✓ Connected\ngbrain-a: ${LOOPBACK} (HTTP) - ✓ Connected`, + }); + expect(runDetect().json.gbrain_mcp_mode).toBe('remote-http'); + }); + + test('tier 2: loopback (SSE) line without a local engine → remote-http', () => { + makeFakeClaude({ hasGetJson: false, hasMcpList: true, mcpListOutput: 'gbrain: http://localhost:3131/sse (SSE) - ✓ Connected' }); + expect(runDetect().json.gbrain_mcp_mode).toBe('remote-http'); + }); + + // detect's two fields must agree for the same loopback entry: local-stdio + // never sits next to thin-client, and remote-http never next to a local + // engine's own failure. + function fakeGbrainSourcesFails(message: string): void { + fs.writeFileSync( + path.join(fakeBinDir, 'gbrain'), + [ + '#!/bin/bash', + 'case "$1" in', + ' --version|version) echo "gbrain 0.60.472.0"; exit 0 ;;', + ` sources) echo "${message}" >&2; exit 1 ;;`, + ' *) exit 0 ;;', + 'esac', + '', + ].join('\n'), + { mode: 0o755 }, + ); + } + + test('a loopback entry over a locked local PGLite reads remote-http + thin-client (Path 4 split-engine)', () => { + fs.mkdirSync(path.join(tmpHome, '.gbrain'), { recursive: true }); + fs.writeFileSync(path.join(tmpHome, '.gbrain', 'config.json'), JSON.stringify({ engine: 'pglite', database_path: '/x/brain.pglite' })); + fs.writeFileSync(path.join(tmpHome, '.claude.json'), JSON.stringify({ mcpServers: { gbrain: { type: 'http', url: LOOPBACK } } })); + fakeGbrainSourcesFails("GBrain's local database is already open through gbrain serve (MCP, PID 4242). This brain uses PGLite"); + const { json } = runDetect(); + expect(json.gbrain_mcp_mode).toBe('remote-http'); + expect(json.gbrain_local_status).toBe('thin-client'); + }); + + test('a loopback entry over a failing local Postgres stays local-stdio and keeps its own status', () => { + localEngine(); + fs.writeFileSync(path.join(tmpHome, '.claude.json'), JSON.stringify({ mcpServers: { gbrain: { type: 'http', url: LOOPBACK } } })); + fakeGbrainSourcesFails('error: connect ECONNREFUSED 127.0.0.1:5432'); + const { json } = runDetect(); + expect(json.gbrain_mcp_mode).toBe('local-stdio'); + expect(json.gbrain_local_status).not.toBe('thin-client'); + }); +}); + describe('gbrain_mcp_mode — no info anywhere', () => { test('no claude binary AND no ~/.claude.json → none', () => { // No fake claude, no file. diff --git a/test/gstack-memory-ingest.test.ts b/test/gstack-memory-ingest.test.ts index dd9df4862..168080d95 100644 --- a/test/gstack-memory-ingest.test.ts +++ b/test/gstack-memory-ingest.test.ts @@ -685,6 +685,101 @@ process.stdout.write(execFileSync(process.execPath, ['-e', 'process.stdout.write expect(staged[0]).not.toContain("_unattributed"); }); + describe("a loopback gbrain serve --http over the local engine is a local brain (#3539)", () => { + const PG = { engine: "postgres", database_url: "postgresql://u@localhost:5432/brain" }; + const loopback = { mcpServers: { gbrain: { type: "http", url: "http://127.0.0.1:3131/mcp", headers: { Authorization: "Bearer x" } } } }; + const remote = { mcpServers: { gbrain: { type: "http", url: "http://fixture.invalid/mcp" } } }; + function gbrainConfig(cfg: Record): void { + mkdirSync(join(home, ".gbrain"), { recursive: true }); + writeFileSync(join(home, ".gbrain", "config.json"), JSON.stringify(cfg)); + } + function claudeJson(cfg: unknown): void { + writeFileSync(join(home, ".claude.json"), JSON.stringify(cfg)); + } + function outgoing(): string[] { + const root = join(env.GSTACK_HOME, "transcripts"); + return existsSync(root) ? readdirSync(root, { recursive: true }).map(String).filter((f: string) => f.endsWith(".md")) : []; + } + // The fixture gbrain answers `import` but not the landing lookup, so an + // imported page stops at imported_unverified; either way it went through + // `gbrain import` into a per-repository transcript source, not the + // remote-http staging. + const importedEntry = { + status: expect.stringMatching(/^(ingested|imported_unverified)$/), + source_id: expect.stringMatching(/^gstack-transcripts-/), + }; + + it("imports through gbrain instead of staging for a brain admin", () => { + scanner("clean"); + gbrainConfig(PG); + claudeJson(loopback); + const path = attributed(source()); + const r = run([]); + expect(r.status, r.stderr).toBe(0); + expect(r.stderr).not.toContain("remote-http mode"); + expect(imported()).toHaveLength(1); + expect(outgoing()).toEqual([]); + expect(sessions()[path]).toMatchObject(importedEntry); + }); + + it("keeps staging for a non-loopback URL even with a local engine", () => { + scanner("clean"); + gbrainConfig(PG); + claudeJson(remote); + const path = attributed(source()); + expect(run([]).status).toBe(0); + expect(imported()).toEqual([]); + expect(outgoing()).toHaveLength(1); + expect(sessions()[path]).toMatchObject({ status: "staged", source_id: "remote-http" }); + }); + + it("keeps staging for a loopback URL over a local PGLite (Path 4 split-engine through a tunnel)", () => { + scanner("clean"); + gbrainConfig({ engine: "pglite", database_path: join(home, ".gbrain", "brain.pglite") }); + claudeJson(loopback); + const path = attributed(source()); + expect(run([]).status).toBe(0); + expect(imported()).toEqual([]); + expect(outgoing()).toHaveLength(1); + expect(sessions()[path]).toMatchObject({ status: "staged", source_id: "remote-http" }); + }); + + it("keeps staging for a loopback URL when gbrain's config carries remote_mcp", () => { + scanner("clean"); + gbrainConfig({ ...PG, remote_mcp: { mcp_url: "http://127.0.0.1:3131/mcp" } }); + claudeJson(loopback); + const path = attributed(source()); + expect(run([]).status).toBe(0); + expect(imported()).toEqual([]); + expect(outgoing()).toHaveLength(1); + expect(sessions()[path]).toMatchObject({ status: "staged", source_id: "remote-http" }); + }); + + const stdio = { mcpServers: { gbrain: { type: "stdio", command: "gbrain", args: ["serve"] } } }; + for (const [label, local] of [["loopback http", loopback], ["stdio", stdio]] as const) { + it(`re-imports pages left staged by an earlier remote-http run once the brain is local (${label})`, () => { + scanner("clean"); + claudeJson(remote); + const path = attributed(source()); + expect(run([]).status).toBe(0); + expect(sessions()[path]).toMatchObject({ status: "staged", source_id: "remote-http" }); + // Still remote: the staged page counts as unchanged, as before. + expect(run(["--probe"]).stdout).toMatch(/Unchanged:\s+1/); + + gbrainConfig(PG); + claudeJson(local); + const probe = run(["--probe"]); + expect(probe.stdout).toMatch(/New \(never ingested\):\s+1/); + expect(probe.stdout).toMatch(/Unchanged:\s+0/); + const r = run(["--incremental"]); + expect(r.status, r.stderr).toBe(0); + expect(imported()).toHaveLength(1); + expect(sessions()[path]).toMatchObject(importedEntry); + expect(run(["--probe"]).stdout).toMatch(/Unchanged:\s+1/); + }); + } + }); + it("scans remote-http pages before persistent staging", () => { scanner("clean"); writeFileSync(join(home, ".claude.json"), JSON.stringify({ mcpServers: { gbrain: { type: "http", url: "http://fixture.invalid/mcp" } } }));