@researai/deepscientist 1.5.17 → 1.6.0

package/src/deepscientist/kimi_cli_compat.py

@@ -0,0 +1,50 @@
+from __future__ import annotations
+
+import shutil
+from pathlib import Path
+
+from .shared import ensure_dir
+
+_RUNTIME_COPY_ALLOWLIST = (
+    "config.toml",
+    "device_id",
+    "credentials",
+    "plugins",
+)
+
+
+def _remove_tree_path(path: Path) -> None:
+    if not path.exists() and not path.is_symlink():
+        return
+    if path.is_symlink() or path.is_file():
+        path.unlink()
+        return
+    shutil.rmtree(path)
+
+
+def materialize_kimi_runtime_home(
+    *,
+    source_home: str | Path,
+    target_home: str | Path,
+) -> Path:
+    source_root = Path(source_home).expanduser()
+    target_root = ensure_dir(Path(target_home))
+    target_kimi_root = target_root / ".kimi"
+
+    if target_kimi_root.exists() or target_kimi_root.is_symlink():
+        _remove_tree_path(target_kimi_root)
+
+    ensure_dir(target_kimi_root)
+    if source_root.exists() and source_root.is_dir():
+        for relative_name in _RUNTIME_COPY_ALLOWLIST:
+            source_path = source_root / relative_name
+            target_path = target_kimi_root / relative_name
+            if not source_path.exists() and not source_path.is_symlink():
+                continue
+            if source_path.is_symlink() or source_path.is_file():
+                ensure_dir(target_path.parent)
+                shutil.copy2(source_path, target_path)
+                continue
+            shutil.copytree(source_path, target_path, dirs_exist_ok=True)
+
+    return target_kimi_root

package/src/deepscientist/latex_runtime.py

@@ -12,7 +12,7 @@ from typing import Any
 from urllib.parse import quote, unquote
 
 from .runtime_tools import RuntimeToolService
-from .shared import ensure_dir, generate_id, resolve_within, utc_now, write_json
+from .shared import ensure_dir, generate_id, resolve_within, utc_now, utf8_text_subprocess_kwargs, write_json
 
 _QUEST_DIR_PREFIX = "quest-dir::"
 _QUEST_FILE_PREFIX = "quest-file::"
@@ -471,8 +471,8 @@ class QuestLatexService:
                 args,
                 cwd=str(cwd),
                 capture_output=True,
-                text=True,
                 check=False,
+                **utf8_text_subprocess_kwargs(),
             )
             header = f"$ {' '.join(args)}\n"
             body = (result.stdout or "") + (result.stderr or "")

package/src/deepscientist/mcp/context.py

@@ -19,6 +19,7 @@ class McpContext:
     worker_id: str | None
     worktree_root: Path | None
     team_mode: str | None
+    custom_profile: str | None = None
 
     @classmethod
     def from_env(cls) -> "McpContext":
@@ -38,6 +39,7 @@ class McpContext:
             worker_id=os.environ.get("DS_WORKER_ID") or None,
             worktree_root=_path("DS_WORKTREE_ROOT"),
             team_mode=os.environ.get("DS_TEAM_MODE") or None,
+            custom_profile=os.environ.get("DS_CUSTOM_PROFILE") or None,
         )
 
     def require_quest_root(self) -> Path:

package/src/deepscientist/mcp/schemas.py

@@ -0,0 +1,114 @@
+from __future__ import annotations
+
+from typing import Any
+
+from pydantic import BaseModel, ConfigDict, Field
+
+
+class MetricEntryPayload(BaseModel):
+    """Per-metric entry inside `metric_contract.metrics`."""
+
+    model_config = ConfigDict(extra="allow")
+
+    metric_id: str = Field(
+        ...,
+        description="Canonical paper-facing metric id, e.g. `accuracy`, `f1`, `rouge_l`.",
+    )
+    description: str | None = Field(
+        default=None,
+        description="Human-readable definition of what the metric measures. Required for the baseline gate.",
+    )
+    derivation: str | None = Field(
+        default=None,
+        description=(
+            "How the metric is derived (formula, aggregation, or processing steps). "
+            "Required unless `origin_path` is provided."
+        ),
+    )
+    origin_path: str | None = Field(
+        default=None,
+        description=(
+            "Dot- or slash-delimited path into `metrics_summary` where the canonical numeric value lives. "
+            "Accepted in place of `derivation`."
+        ),
+    )
+    source_ref: str | None = Field(
+        default=None,
+        description=(
+            "Canonical source reference (paper section, leaderboard URL, repo path, or commit ref). "
+            "Required for the baseline gate."
+        ),
+    )
+    direction: str | None = Field(
+        default=None,
+        description="`higher_is_better` or `lower_is_better`.",
+    )
+    label: str | None = Field(default=None, description="Display label; defaults to `metric_id`.")
+    unit: str | None = Field(default=None, description="Unit, e.g. `%` or `ms`.")
+    decimals: int | None = Field(default=None, description="Preferred decimal places for display.")
+    chart_group: str | None = Field(default=None, description="Optional chart grouping key.")
+    required: bool | None = Field(
+        default=None,
+        description="Whether downstream experiments must cover this metric. Defaults to true.",
+    )
+
+
+class PrimaryMetricPayload(BaseModel):
+    """Headline gate / scoreboard metric."""
+
+    model_config = ConfigDict(extra="allow")
+
+    metric_id: str = Field(
+        ...,
+        description="Metric id; must match one of the entries in `metric_contract.metrics`.",
+    )
+    value: float | None = Field(default=None, description="Numeric baseline value for the headline metric.")
+    direction: str | None = Field(default=None, description="`higher_is_better` or `lower_is_better`.")
+
+
+class MetricContractPayload(BaseModel):
+    """Canonical comparison contract submitted with `confirm_baseline`."""
+
+    model_config = ConfigDict(extra="allow")
+
+    primary_metric_id: str | None = Field(
+        default=None,
+        description="Id of the headline metric; must match one of the entries in `metrics`.",
+    )
+    contract_id: str | None = Field(
+        default=None,
+        description="Stable contract identifier; defaults to the baseline id.",
+    )
+    metrics: list[MetricEntryPayload] = Field(
+        default_factory=list,
+        description=(
+            "Per-metric entries with `description`, `derivation` or `origin_path`, and `source_ref`. "
+            "Descriptions nested inside `metrics_summary`, `primary_metric`, or "
+            "`json/metric_contract.json` are not honored."
+        ),
+    )
+    evaluation_protocol: dict[str, Any] | None = Field(
+        default=None,
+        description="Optional canonical evaluation protocol (scope_id, code_paths, code_hashes, ...).",
+    )
+
+
+class SupplementaryBaselinePayload(BaseModel):
+    """Supplementary baseline entry used by overwrite_baseline / analysis inventory refresh."""
+
+    model_config = ConfigDict(extra="allow")
+
+    baseline_id: str = Field(..., description="Stable baseline id.")
+    variant_id: str | None = Field(default=None, description="Optional baseline variant id.")
+    reason: str | None = Field(default=None, description="Why this supplementary baseline matters now.")
+    benchmark: str | None = Field(default=None, description="Optional benchmark label.")
+    split: str | None = Field(default=None, description="Optional split label.")
+    baseline_root_rel_path: str | None = Field(
+        default=None,
+        description="Quest-relative baseline root under `baselines/local/...` or `baselines/imported/...`.",
+    )
+    metrics_summary: dict[str, Any] | None = Field(default=None, description="Flat canonical metrics summary.")
+    evidence_paths: list[str] | None = Field(default=None, description="Optional supporting evidence paths.")
+    published: bool | None = Field(default=None, description="Whether this supplementary baseline was published.")
+    published_entry_id: str | None = Field(default=None, description="Published registry entry id when applicable.")
+    status: str | None = Field(default=None, description="Supplementary baseline status.")