loki-mode 11.26.0 → 11.28.0

package/SKILL.md

@@ -3,7 +3,7 @@ name: loki-mode
 description: An autonomous software factory that knows what it is supposed to deliver, and proves it did. Use when the user says Loki Mode or asks to build, fix or verify software autonomously.
 ---
 
-# Loki Mode v11.26.0
+# Loki Mode v11.28.0
 
 **Current entry points (Loki 10 engine):** `loki "<task>"`, `loki owner/repo#N` and `loki quick "<task>"` run the Loki 10 engine and end with a signed receipt (`loki verify`). `loki backlog owner/repo --all` runs many issues in parallel. Guide: `docs/v10/GUIDE.md`. The skill instructions below, and `loki start`, describe the previous bash loop. PRD-file starts (`loki start ./prd.md`) and the opencode provider still run on it until FC38-SWEEP ports them; the legacy-engine command and the LOKI_ENGINE switch remain removed. Prefer `loki owner/repo#N` for issues.
 
@@ -496,4 +496,4 @@ See `CHANGELOG.md` entries [7.5.7], [7.5.8], [7.5.13] for the per-fix list and r
 
 ---
 
-**v11.26.0 | [Autonomi](https://www.autonomi.dev/) flagship product | core skill**
+**v11.28.0 | [Autonomi](https://www.autonomi.dev/) flagship product | core skill**

package/VERSION

@@ -1 +1 @@
-11.26.0
+11.28.0

package/autonomy/context-tracker.py

@@ -12,7 +12,7 @@ Provider-agnostic: supports Claude, Codex, and Gemini CLIs.
 Called by run.sh after each RARV iteration.
 
 Usage:
-    python3 context-tracker.py --iteration N --loki-dir .loki [--window-size 200000]
+    python3 context-tracker.py --iteration N --loki-dir .loki [--window-size N] [--model ID]
     python3 context-tracker.py --iteration N --loki-dir .loki --provider codex \
         --tokens-input 5000 --tokens-output 2000
     python3 context-tracker.py --iteration N --loki-dir .loki --session-file /path/to/session.jsonl
@@ -57,12 +57,29 @@ PRICING_BY_PROVIDER = {
     },
 }
 
-# Default context window sizes per provider
-WINDOW_SIZES = {
-    "claude": 200000,
+# Last-resort window for providers whose models carry no context_window in
+# providers/model_catalog.json (the catalog owns Claude model capacity).
+FALLBACK_WINDOW_SIZES = {
     "codex": 200000,
     "gemini": 1000000,
 }
+DEFAULT_WINDOW_SIZE = 200000
+CATALOG_PATH = Path(__file__).resolve().parent.parent / "providers" / "model_catalog.json"
+
+
+def resolve_window_size(provider, model=None):
+    """Context window in tokens: catalog by model id, else the provider's
+    catalog default model, else FALLBACK_WINDOW_SIZES."""
+    try:
+        entry = json.loads(CATALOG_PATH.read_text())["providers"].get(provider, {})
+        models = {m.get("id"): m for m in entry.get("models", []) if isinstance(m, dict)}
+        for mid in (model, entry.get("latest_development")):
+            window = models.get(mid, {}).get("context_window")
+            if isinstance(window, int) and window > 0:
+                return window
+    except (OSError, ValueError, KeyError, AttributeError):
+        pass
+    return FALLBACK_WINDOW_SIZES.get(provider, DEFAULT_WINDOW_SIZE)
 
 
 def get_pricing(provider):
@@ -153,6 +170,8 @@ def parse_session(jsonl_path, last_offset, iteration, window_size):
     batch_cache_read = 0
     batch_cache_creation = 0
 
+    latest_model = None
+
     # Track latest cumulative usage (last assistant message has running total for that API call)
     latest_input = 0
     latest_output = 0
@@ -189,6 +208,8 @@ def parse_session(jsonl_path, last_offset, iteration, window_size):
             # Extract token usage from assistant messages
             if entry_type == "assistant":
                 usage = entry.get("message", {}).get("usage", {})
+                if entry.get("message", {}).get("model"):
+                    latest_model = entry["message"]["model"]
                 if usage:
                     inp = usage.get("input_tokens", 0)
                     out = usage.get("output_tokens", 0)
@@ -213,19 +234,21 @@ def parse_session(jsonl_path, last_offset, iteration, window_size):
         "batch_cache_creation": batch_cache_creation,
         "latest_input": latest_input,
         "latest_cache_read": latest_cache_read,
+        "model": latest_model,
         "compactions": compactions,
         "new_offset": current_line,
     }
 
 
-def update_tracking(loki_dir, iteration, window_size, provider="claude",
+def update_tracking(loki_dir, iteration, window_size, provider="claude", model=None,
                     session_file=None, direct_tokens=None):
     """Main tracking function - parse session and update tracking.json.
 
     Args:
         loki_dir: Path to .loki directory
         iteration: Current RARV iteration number
-        window_size: Context window size in tokens
+        window_size: Context window size in tokens (0 = resolve from the catalog)
+        model: Run model id used to look up the window (default: model seen in the session)
         provider: Provider name (claude, codex, gemini)
         session_file: Explicit path to session file (optional)
         direct_tokens: Dict with input/output token counts for non-Claude providers
@@ -315,6 +338,9 @@ def update_tracking(loki_dir, iteration, window_size, provider="claude",
         tracking["updated_at"] = datetime.now(timezone.utc).isoformat()
         offset_file.write_text(str(result["new_offset"]))
 
+    if window_size <= 0:
+        window_size = resolve_window_size(provider, model or result.get("model"))
+
     # Calculate iteration cost
     iter_cost = calculate_cost(
         result["batch_input"],
@@ -379,6 +405,8 @@ def main():
                         help="Context window size in tokens (0 = auto-detect from provider)")
     parser.add_argument("--provider", default="claude", choices=["claude", "codex", "gemini"],
                         help="LLM provider (claude, codex, gemini)")
+    parser.add_argument("--model", default=None,
+                        help="Run model id for the catalog window lookup (default: model seen in the session)")
     parser.add_argument("--session-file", default=None,
                         help="Explicit path to session file (overrides auto-detection)")
     parser.add_argument("--tokens-input", type=int, default=0,
@@ -387,8 +415,8 @@ def main():
                         help="Direct output token count (for non-Claude providers)")
     args = parser.parse_args()
 
-    # Auto-detect window size from provider if not specified
-    window_size = args.window_size if args.window_size > 0 else WINDOW_SIZES.get(args.provider, 200000)
+    # window_size 0 is resolved from the catalog once the run's model is known
+    window_size = max(args.window_size, 0)
 
     # Build direct tokens dict if provided
     direct_tokens = None
@@ -396,7 +424,7 @@ def main():
         direct_tokens = {"input": args.tokens_input, "output": args.tokens_output}
 
     update_tracking(args.loki_dir, args.iteration, window_size,
-                    provider=args.provider, session_file=args.session_file,
+                    provider=args.provider, model=args.model, session_file=args.session_file,
                     direct_tokens=direct_tokens)
 
 

package/dashboard/__init__.py

@@ -7,6 +7,6 @@ Modules:
     control: Session control helpers (status, atomic writes, origin check)
 """
 
-__version__ = "11.26.0"
+__version__ = "11.28.0"
 
 __all__ = ["__version__"]

package/docs/CLI-REFERENCE.md

@@ -1,6 +1,6 @@
 # CLI reference
 
-Generated from `loki-ts/src/cli/registry.ts` by `scripts/generate-stale-zero.sh` for v11.26.0. Do not edit by hand.
+Generated from `loki-ts/src/cli/registry.ts` by `scripts/generate-stale-zero.sh` for v11.28.0. Do not edit by hand.
 Hidden and legacy-only commands are omitted; see `docs/v10/CLI-MODERN.md` for the full inventory.
 
 | Command | Description |

package/docs/INSTALLATION.md

@@ -2,7 +2,7 @@
 
 Loki Mode, by [Autonomi](https://www.autonomi.dev/), turns a spec or an issue into a verified pull request with a receipt you can check. This page covers installing it, a first run, and the options around it.
 
-**Version:** v11.26.0
+**Version:** v11.28.0
 
 **Engine note:** `loki start` is the primary command. `loki start`, `loki "<task>"`, `loki owner/repo#N` and `loki quick` all run the Loki 10 engine, whatever the flag order or input form (PRD file, issue ref, task or one-word brief). The previous loop runs only on `--engine legacy`, `LOKI_ENGINE=legacy`, or for one call when a flag Loki 10 cannot honor is present. See the [user guide](v10/GUIDE.md) and the [5-minute quickstart](QUICKSTART.md).