@vectorize-io/hindsight-coding-agents 0.5.4 → 0.6.0
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/README.md +24 -7
- package/dist/antigravity-hook.js +205 -81
- package/dist/antigravity-statusline.js +32 -22
- package/dist/antigravity-stop-hook.js +216 -79
- package/dist/claude-hook.js +205 -81
- package/dist/claude-sessionstart-hook.js +167 -105
- package/dist/claude-stop-hook.js +216 -79
- package/dist/cline.js +279 -86
- package/dist/codex-hook.js +205 -81
- package/dist/codex-sessionstart-hook.js +167 -105
- package/dist/codex-stop-hook.js +216 -79
- package/dist/copilot-hook.js +205 -81
- package/dist/copilot-sessionstart-hook.js +167 -105
- package/dist/copilot-stop-hook.js +216 -79
- package/dist/cursor-hook.js +205 -81
- package/dist/cursor-sessionstart-hook.js +167 -105
- package/dist/cursor-stop-hook.js +216 -79
- package/dist/daemon-start.js +28 -18
- package/dist/dcode-hook.js +205 -81
- package/dist/dcode-sessionstart-hook.js +167 -105
- package/dist/dcode-stop-hook.js +216 -79
- package/dist/deepen.js +98 -40
- package/dist/devin-hook.js +205 -81
- package/dist/devin-sessionstart-hook.js +167 -105
- package/dist/devin-stop-hook.js +216 -79
- package/dist/droid-hook.js +205 -81
- package/dist/droid-sessionstart-hook.js +167 -105
- package/dist/droid-stop-hook.js +216 -79
- package/dist/dsh.js +277 -85
- package/dist/grok-hook.js +205 -81
- package/dist/grok-sessionstart-hook.js +167 -105
- package/dist/grok-stop-hook.js +216 -79
- package/dist/hindsight-seed.js +32 -22
- package/dist/index.js +280 -88
- package/dist/installer.js +308 -214
- package/dist/kilo.js +280 -88
- package/dist/mcp-server.js +97 -39
- package/dist/opencode2.js +279 -87
- package/dist/pi.js +279 -87
- package/dist/prime-agent.js +279 -87
- package/dist/qwen-hook.js +205 -81
- package/dist/qwen-sessionstart-hook.js +167 -105
- package/dist/qwen-stop-hook.js +216 -79
- package/dist/status.js +94 -36
- package/dist/zcode-hook.js +205 -81
- package/dist/zcode-sessionstart-hook.js +167 -105
- package/dist/zcode-stop-hook.js +216 -79
- package/package.json +1 -1
- package/plugin.json +1 -1
- package/skill/SKILL.md +24 -7
package/skill/SKILL.md
CHANGED
|
@@ -69,6 +69,7 @@ npx @vectorize-io/hindsight-coding-agents install all # every detected
|
|
|
69
69
|
npx @vectorize-io/hindsight-coding-agents install claude-code # or just one
|
|
70
70
|
npx @vectorize-io/hindsight-coding-agents uninstall all # removes exactly what install added
|
|
71
71
|
npx @vectorize-io/hindsight-coding-agents update # refresh the runtime only, no rewiring
|
|
72
|
+
npx @vectorize-io/hindsight-coding-agents stats # how often each agent uses Hindsight
|
|
72
73
|
```
|
|
73
74
|
|
|
74
75
|
`install` takes an explicit target — `all`, or one or more harness names. A bare
|
|
@@ -121,8 +122,8 @@ as `HINDSIGHT_MAX_PARALLEL_RETAINS` for containers and CI.
|
|
|
121
122
|
|
|
122
123
|
`HINDSIGHT_CONFIG` moves the file itself — point it at another path for a container or a test
|
|
123
124
|
harness where `$HOME` is not the right anchor. It is still exactly one file; only its location
|
|
124
|
-
changes. (The other variables that are not settings are `HINDSIGHT_LOG_FILE`, `HINDSIGHT_DIAG_FILE
|
|
125
|
-
and `HINDSIGHT_LOG_LEVEL` — see [Diagnostics & logging](#diagnostics--logging).)
|
|
125
|
+
changes. (The other variables that are not settings are `HINDSIGHT_LOG_FILE`, `HINDSIGHT_DIAG_FILE`,
|
|
126
|
+
`HINDSIGHT_USAGE_FILE` and `HINDSIGHT_LOG_LEVEL` — see [Diagnostics & logging](#diagnostics--logging).)
|
|
126
127
|
|
|
127
128
|
### When a change takes effect
|
|
128
129
|
|
|
@@ -204,7 +205,7 @@ hook by Codex...), so one shared config serves several agents side by side:
|
|
|
204
205
|
| `manageBankConfig` | `true` | let the plugin shape the bank's own configuration — the retain strategies it writes under, the `knowledge` entity-label group, and, on a bank that has none, the missions. Writing is strictly **additive**: it adds what the bank does not define and never overwrites what is there, so your control-plane edits survive. Set `false` to keep it out of the bank config entirely — see **A bank you shape yourself** below |
|
|
205
206
|
| `observationScopes` | `"shared"` | how consolidation groups observations: `"shared"` (default) = ONE global scope per bank, so every agent on a repo builds one set of beliefs; also `"combined"` (the server default), `"per_tag"`, `"all_combinations"`, `[["t"]]`; `"per_source"` adds a scope per `source:` kind alongside the global one, so commit knowledge and conversation knowledge consolidate apart |
|
|
206
207
|
| `disabled` | `false` | hard off-switch (inert plugin/hook — a no-memory baseline) |
|
|
207
|
-
| `reflectTimeoutMs` | `120000` | **automatic** session-reflect timeout (hook harnesses additionally cap it at
|
|
208
|
+
| `reflectTimeoutMs` | `120000` | **automatic** session-reflect timeout (hook harnesses additionally cap it at 20s to fit the host's hook window); on timeout or a 5xx the hook falls back to knowledge-page search, then to a raw recall of observations (recorded) |
|
|
208
209
|
| `reflectToolTimeoutMs` | `330000` | timeout for the agent-invoked `hindsight_reflect` tool — a call the agent waits on, whose high-budget synthesis on a populated bank runs for minutes. Defaults above the server's own reflect wall timeout (`HINDSIGHT_API_REFLECT_WALL_TIMEOUT`, 300s) so the server decides when to give up. Unset, it inherits an explicitly raised `reflectTimeoutMs`, but a short one never lowers it |
|
|
209
210
|
| `reflectBudget` | `"high"` | reflect budget for the `hindsight_reflect` tool: `"low"`, `"mid"` or `"high"`. Drop it on a large bank where high-budget synthesis exceeds the server's wall timeout. The automatic session-start reflect always uses `"low"` to fit its hook window and is unaffected |
|
|
210
211
|
| `autoReflect` | `true` | inject a one-time reflect synthesis on the session's **first prompt**. `false` = tool-only reflect: nothing is injected; the agent searches knowledge pages first and reflects only when they are too shallow |
|
|
@@ -459,16 +460,18 @@ they stay where they were built, and new work accrues under the new setting.
|
|
|
459
460
|
|
|
460
461
|
## Diagnostics & logging
|
|
461
462
|
|
|
462
|
-
|
|
463
|
+
All logs live in `~/.hindsight/coding-agents-logs/` (owner-only). Each file rotates to `<file>.1`
|
|
464
|
+
at 10 MB.
|
|
463
465
|
|
|
464
|
-
**Leveled plugin log** (humans debugging):
|
|
466
|
+
**Leveled plugin log** (humans debugging): `~/.hindsight/coding-agents-logs/plugin.log` (override
|
|
465
467
|
`HINDSIGHT_LOG_FILE`) — timestamped `LEVEL [scope] message` lines from every component, including
|
|
466
468
|
the ingestion engine. Level defaults to `info`; set `"logLevel": "debug"` in config or
|
|
467
469
|
`HINDSIGHT_LOG_LEVEL=debug` for ad-hoc debugging (at `debug`, every diag event below is mirrored
|
|
468
470
|
here too, so one file tells the whole story).
|
|
469
471
|
|
|
470
472
|
**Structured diag events** (machines/harnesses): every reflect and page-fetch outcome is appended
|
|
471
|
-
as a JSON line to
|
|
473
|
+
as a JSON line to `~/.hindsight/coding-agents-logs/diag.jsonl` (override with
|
|
474
|
+
`HINDSIGHT_DIAG_FILE`):
|
|
472
475
|
|
|
473
476
|
```json
|
|
474
477
|
{
|
|
@@ -482,9 +485,23 @@ as a JSON line to `/tmp/hindsight-plugin.log` (override with `HINDSIGHT_DIAG_FIL
|
|
|
482
485
|
```
|
|
483
486
|
|
|
484
487
|
`reflect_failed` / `pages_failed` record the error; if you're comparing memory-on vs memory-off,
|
|
485
|
-
check this file — a run whose reflects failed is a no-memory run.
|
|
488
|
+
check this file — a run whose reflects failed is a no-memory run. When the failure was a timeout or
|
|
489
|
+
a 5xx, the hook falls back to knowledge-page search and, if no page matches, to a raw recall of the
|
|
490
|
+
bank's observations: `reflect_fallback_pages` / `reflect_fallback_observations` record what each
|
|
491
|
+
step returned (`*_failed` when it errored). Seed starts are logged as
|
|
486
492
|
`seed_started`.
|
|
487
493
|
|
|
494
|
+
**Tool usage** (is the agent using Hindsight?): one JSON line per finished user turn in
|
|
495
|
+
`~/.hindsight/coding-agents-logs/usage.jsonl` (override `HINDSIGHT_USAGE_FILE`) — the `hindsight_*`
|
|
496
|
+
tools the agent called during that turn, and whether its reply credited Hindsight memory ("From
|
|
497
|
+
Hindsight memory"). The credit rate counts only turns that called a retrieval tool (search, list,
|
|
498
|
+
read, reflect); saving a document is not expected to be credited. Recorded when the session is written back, so a scope with
|
|
499
|
+
`retainSessions: false` records none. It never leaves your machine. Summarize it per agent with:
|
|
500
|
+
|
|
501
|
+
```bash
|
|
502
|
+
npx @vectorize-io/hindsight-coding-agents stats
|
|
503
|
+
```
|
|
504
|
+
|
|
488
505
|
### Is the memory ready yet?
|
|
489
506
|
|
|
490
507
|
`hindsight_sync_status` — the agent-facing tool, `dist/status.js` for scripts — answers exactly
|