gemcode 0.3.69__tar.gz → 0.3.71__tar.gz

{gemcode-0.3.69/src/gemcode.egg-info → gemcode-0.3.71}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: gemcode
-Version: 0.3.69
+Version: 0.3.71
 Summary: Local-first coding agent on Google Gemini + ADK
 Author: GemCode Contributors
 License:                                  Apache License
@@ -217,6 +217,16 @@ gemcode --yes "Add a module docstring to src/foo.py"
 gemcode --session mysess --yes "Continue: run tests and fix failures"
 ```
 
+### What GemCode writes to `.gemcode/`
+
+GemCode keeps project-local state under `.gemcode/`:
+
+- **`sessions.sqlite`**: session events/history (ADK `SqliteSessionService`)
+- **`audit.log`**: JSONL audit trail for tool usage + model usage + stop reasons
+- **`tool-results/`**: oversized tool outputs offloaded to stable refs (`tool_result:<sha>`)
+- **`artifacts/`**: file artifacts (ADK `FileArtifactService`)
+- **`policy.json`**: self-tuning per-repo profile used to calibrate dynamic budgets
+
 - **`--yes`**: allow mutating tools (`write_file`, `search_replace`). Shell execution is still restricted by the `.env.example` allowlist.
 - **`--session`**: Conversation history is stored under `.gemcode/sessions.sqlite` (ADK `SqliteSessionService`). Reuse the same `--session` id to continue.
 - **`--max-llm-calls`**: cap model↔tool iterations for this message (maps to ADK `RunConfig.max_llm_calls`). You can also set `GEMCODE_MAX_LLM_CALLS`.
@@ -243,6 +253,33 @@ gemcode --session mysess --yes "Continue: run tests and fix failures"
 - **Recovery-loop**: ADK `ReflectAndRetryToolPlugin`-based retries on tool failures.
   - Set `GEMCODE_ENABLE_TOOL_RECOVERY_RETRY=0` to disable.
   - Set `GEMCODE_TOOL_REFLECT_MAX_RETRIES=1` to control retries per tool.
+
+### Token efficiency (dynamic + intelligent)
+
+GemCode optimizes tokens without losing capability by using a dynamic policy:
+
+- **Context-pressure aware**: tool caps tighten when context is tight, loosen when there is room.
+- **Risk/complexity aware**: caps increase for risky tasks (writes, shell, failures, many files).
+- **Self-tuning per repo**: `.gemcode/policy.json` calibrates baseline evidence budgets over time.
+
+Key env toggles:
+
+- `GEMCODE_DYNAMIC_TOKEN_POLICY=0|1`
+- `GEMCODE_DYNAMIC_RISK_POLICY=0|1`
+- `GEMCODE_DYNAMIC_RISK_BOOST=<float>` (default `0.6`)
+- `GEMCODE_TOOL_RESULT_OFFLOAD=0|1` (default `1`)
+
+Live telemetry:
+
+- `/status` shows `risk_score`, `context_percent_left`, and profile EMAs.
+
+### Stable tool output offloading (OpenClaude-style)
+
+Oversized tool outputs are automatically offloaded and replaced with stable refs:
+
+- Stored in: `.gemcode/tool-results/`
+- References: `tool_result:<sha256>`
+- Load on demand: `load_tool_result(ref)`
 - **Gemini thinking controls (Claude-like)**:
   - By default GemCode lets Gemini use its dynamic/adaptive thinking behavior.
   - Set `GEMCODE_DISABLE_THINKING=1` to force a best-effort “low thinking” mode:
@@ -301,12 +338,24 @@ the user’s project.
   - `list_directory`
   - `glob_files`
   - `grep_content`
+  - `repo_map`
+  - `web_search`
+  - `web_fetch`
+  - `notebook_read`
+  - `notebook_edit`
 - Mutating tools (require `--yes` unless your policy blocks them):
   - `write_file`
   - `search_replace`
 - Shell execution:
   - `run_command` (guarded by `GEMCODE_ALLOW_COMMANDS` from `.env.example` and
     `GEMCODE_PERMISSION_MODE`).
+  - `bash` (pipelines + redirects; supports `background=True`)
+
+- Background task management (for processes started via `bash(..., background=True)`):
+  - `list_tasks`, `task_output`, `kill_task`
+
+- Tool offload loader:
+  - `load_tool_result(ref)`
 
 Tool execution is still controlled by permission gates and then governed by
 GemCode’s circuit breaker + recovery behavior.
@@ -434,6 +483,23 @@ pip install -e ".[dev]"
 pytest
 ```
 
+## Release workflow (GitHub Actions → PyPI)
+
+This repository publishes to PyPI on tag pushes:
+
+- `.github/workflows/publish-pypi.yml` triggers on tags matching `v*`
+
+Typical release steps:
+
+```bash
+# 1) bump gemcode/pyproject.toml version
+git add -A
+git commit -m "release: vX.Y.Z"
+git tag -a vX.Y.Z -m "vX.Y.Z"
+git push origin HEAD
+git push origin vX.Y.Z
+```
+
 ## References (local only)
 
 Do not commit proprietary leaked trees into this package. Keep `claude-code-leaked/` and similar folders outside version control or in a private mirror.

{gemcode-0.3.69 → gemcode-0.3.71}/README.md

@@ -28,6 +28,16 @@ gemcode --yes "Add a module docstring to src/foo.py"
 gemcode --session mysess --yes "Continue: run tests and fix failures"
 ```
 
+### What GemCode writes to `.gemcode/`
+
+GemCode keeps project-local state under `.gemcode/`:
+
+- **`sessions.sqlite`**: session events/history (ADK `SqliteSessionService`)
+- **`audit.log`**: JSONL audit trail for tool usage + model usage + stop reasons
+- **`tool-results/`**: oversized tool outputs offloaded to stable refs (`tool_result:<sha>`)
+- **`artifacts/`**: file artifacts (ADK `FileArtifactService`)
+- **`policy.json`**: self-tuning per-repo profile used to calibrate dynamic budgets
+
 - **`--yes`**: allow mutating tools (`write_file`, `search_replace`). Shell execution is still restricted by the `.env.example` allowlist.
 - **`--session`**: Conversation history is stored under `.gemcode/sessions.sqlite` (ADK `SqliteSessionService`). Reuse the same `--session` id to continue.
 - **`--max-llm-calls`**: cap model↔tool iterations for this message (maps to ADK `RunConfig.max_llm_calls`). You can also set `GEMCODE_MAX_LLM_CALLS`.
@@ -54,6 +64,33 @@ gemcode --session mysess --yes "Continue: run tests and fix failures"
 - **Recovery-loop**: ADK `ReflectAndRetryToolPlugin`-based retries on tool failures.
   - Set `GEMCODE_ENABLE_TOOL_RECOVERY_RETRY=0` to disable.
   - Set `GEMCODE_TOOL_REFLECT_MAX_RETRIES=1` to control retries per tool.
+
+### Token efficiency (dynamic + intelligent)
+
+GemCode optimizes tokens without losing capability by using a dynamic policy:
+
+- **Context-pressure aware**: tool caps tighten when context is tight, loosen when there is room.
+- **Risk/complexity aware**: caps increase for risky tasks (writes, shell, failures, many files).
+- **Self-tuning per repo**: `.gemcode/policy.json` calibrates baseline evidence budgets over time.
+
+Key env toggles:
+
+- `GEMCODE_DYNAMIC_TOKEN_POLICY=0|1`
+- `GEMCODE_DYNAMIC_RISK_POLICY=0|1`
+- `GEMCODE_DYNAMIC_RISK_BOOST=<float>` (default `0.6`)
+- `GEMCODE_TOOL_RESULT_OFFLOAD=0|1` (default `1`)
+
+Live telemetry:
+
+- `/status` shows `risk_score`, `context_percent_left`, and profile EMAs.
+
+### Stable tool output offloading (OpenClaude-style)
+
+Oversized tool outputs are automatically offloaded and replaced with stable refs:
+
+- Stored in: `.gemcode/tool-results/`
+- References: `tool_result:<sha256>`
+- Load on demand: `load_tool_result(ref)`
 - **Gemini thinking controls (Claude-like)**:
   - By default GemCode lets Gemini use its dynamic/adaptive thinking behavior.
   - Set `GEMCODE_DISABLE_THINKING=1` to force a best-effort “low thinking” mode:
@@ -112,12 +149,24 @@ the user’s project.
   - `list_directory`
   - `glob_files`
   - `grep_content`
+  - `repo_map`
+  - `web_search`
+  - `web_fetch`
+  - `notebook_read`
+  - `notebook_edit`
 - Mutating tools (require `--yes` unless your policy blocks them):
   - `write_file`
   - `search_replace`
 - Shell execution:
   - `run_command` (guarded by `GEMCODE_ALLOW_COMMANDS` from `.env.example` and
     `GEMCODE_PERMISSION_MODE`).
+  - `bash` (pipelines + redirects; supports `background=True`)
+
+- Background task management (for processes started via `bash(..., background=True)`):
+  - `list_tasks`, `task_output`, `kill_task`
+
+- Tool offload loader:
+  - `load_tool_result(ref)`
 
 Tool execution is still controlled by permission gates and then governed by
 GemCode’s circuit breaker + recovery behavior.
@@ -245,6 +294,23 @@ pip install -e ".[dev]"
 pytest
 ```
 
+## Release workflow (GitHub Actions → PyPI)
+
+This repository publishes to PyPI on tag pushes:
+
+- `.github/workflows/publish-pypi.yml` triggers on tags matching `v*`
+
+Typical release steps:
+
+```bash
+# 1) bump gemcode/pyproject.toml version
+git add -A
+git commit -m "release: vX.Y.Z"
+git tag -a vX.Y.Z -m "vX.Y.Z"
+git push origin HEAD
+git push origin vX.Y.Z
+```
+
 ## References (local only)
 
 Do not commit proprietary leaked trees into this package. Keep `claude-code-leaked/` and similar folders outside version control or in a private mirror.

{gemcode-0.3.69 → gemcode-0.3.71}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "gemcode"
-version = "0.3.69"
+version = "0.3.71"
 description = "Local-first coding agent on Google Gemini + ADK"
 readme = "README.md"
 requires-python = ">=3.11"

{gemcode-0.3.69 → gemcode-0.3.71}/src/gemcode/callbacks.py

@@ -45,6 +45,12 @@ _LAST_CONTEXT_PCT = "gemcode:last_context_percent_left"
 _LAST_CONTEXT_LEVEL = "gemcode:last_context_alert_level"
 _RISK_FILES_TOUCHED = "gemcode:risk_files_touched"
 _RISK_TOOL_CALLS = "gemcode:risk_tool_calls"
+_RISK_HAD_SHELL = "gemcode:risk_had_shell"
+_RISK_HAD_WRITE = "gemcode:risk_had_write"
+_RISK_HAD_FAILURE = "gemcode:risk_had_failure"
+_TOOL_GROUP_CHARS = "gemcode:tool_group_chars"
+_TOOL_GROUP_EXCEEDED = "gemcode:tool_group_budget_exceeded"
+_TOOL_SEQ = "gemcode:tool_seq"
 
 def _truthy_env(name: str, *, default: bool = False) -> bool:
   v = os.environ.get(name)
@@ -147,6 +153,8 @@ def make_before_tool_callback(cfg: GemCodeConfig):
     try:
       if tool_context is not None:
         st = tool_context.state
+        # Per-turn tool sequence (used for stable tool-result replacement keys).
+        st[_TOOL_SEQ] = int(st.get(_TOOL_SEQ, 0) or 0) + 1
         st[_RISK_TOOL_CALLS] = int(st.get(_RISK_TOOL_CALLS, 0) or 0) + 1
         if name == "read_file":
           p = (args or {}).get("path")
@@ -165,9 +173,11 @@ def make_before_tool_callback(cfg: GemCodeConfig):
             object.__setattr__(cfg, "_risk_score", cur)
         # Writes / shell are inherently higher risk; allow more evidence.
         if name in MUTATING_TOOLS:
+          st[_RISK_HAD_WRITE] = True
           cur = float(getattr(cfg, "_risk_score", 0.0) or 0.0)
           object.__setattr__(cfg, "_risk_score", min(1.0, cur + 0.12))
         if name in SHELL_TOOLS:
+          st[_RISK_HAD_SHELL] = True
           cur = float(getattr(cfg, "_risk_score", 0.0) or 0.0)
           object.__setattr__(cfg, "_risk_score", min(1.0, cur + 0.08))
     except Exception:
@@ -355,18 +365,40 @@ def make_after_tool_callback(cfg: GemCodeConfig):
     except Exception:
       pass
 
+    # Aggregate per-turn tool-result budget: if we already exceeded the budget,
+    # tighten caps further for the rest of this user message.
+    try:
+      if tool_context is not None:
+        st = tool_context.state
+        if bool(st.get(_TOOL_GROUP_EXCEEDED, False)):
+          effective_tool_chars = max(1500, int(effective_tool_chars * 0.5))
+    except Exception:
+      pass
+
     if (
       isinstance(tool_response, dict)
       and getattr(cfg, "tool_result_offload_enabled", False)
       and effective_tool_chars > 0
     ):
       try:
-        from gemcode.tool_result_store import maybe_offload_tool_result
-        new_payload, did = maybe_offload_tool_result(
+        from gemcode.tool_result_store import maybe_offload_tool_result_stable
+        seq = None
+        st = None
+        try:
+          if tool_context is not None:
+            st = tool_context.state
+            seq = int(st.get(_TOOL_SEQ, 0) or 0)
+        except Exception:
+          st = None
+          seq = None
+        new_payload, did = maybe_offload_tool_result_stable(
           project_root=cfg.project_root,
           tool_name=name,
+          args=args or {},
           payload=tool_response,
           max_inline_chars=int(effective_tool_chars),
+          state=st,
+          seq=seq,
         )
         if did and isinstance(new_payload, dict):
           tool_response = new_payload
@@ -387,6 +419,19 @@ def make_after_tool_callback(cfg: GemCodeConfig):
       st = tool_context.state
     except Exception:
       return tool_response if (truncated or offloaded) else None
+
+    # Update aggregate per-turn budget counters (best-effort).
+    try:
+      from gemcode.context_budget import estimate_obj_string_chars
+      budget = int(getattr(cfg, "tool_result_group_budget_chars", 0) or 0)
+      if budget > 0:
+        used = int(st.get(_TOOL_GROUP_CHARS, 0) or 0)
+        used += int(estimate_obj_string_chars(tool_response))
+        st[_TOOL_GROUP_CHARS] = used
+        if used >= budget:
+          st[_TOOL_GROUP_EXCEEDED] = True
+    except Exception:
+      pass
     err = isinstance(tool_response, dict) and tool_response.get("error")
     err_kind = (
       isinstance(tool_response, dict) and tool_response.get("error_kind")
@@ -423,9 +468,17 @@ def make_after_tool_callback(cfg: GemCodeConfig):
       cur = float(getattr(cfg, "_risk_score", 0.0) or 0.0)
       bump = 0.0
       if err:
+        try:
+          st[_RISK_HAD_FAILURE] = True
+        except Exception:
+          pass
         bump += 0.15
       if isinstance(tool_response, dict) and isinstance(tool_response.get("exit_code"), int):
         if int(tool_response["exit_code"]) != 0:
+          try:
+            st[_RISK_HAD_FAILURE] = True
+          except Exception:
+            pass
           bump += 0.10
           # Test/build failures should boost evidence allowance more.
           if name in ("bash", "run_command"):
@@ -438,6 +491,27 @@ def make_after_tool_callback(cfg: GemCodeConfig):
       object.__setattr__(cfg, "_risk_score", cur)
     except Exception:
       pass
+
+    # Persist repo calibration profile (best-effort).
+    try:
+      files = st.get(_RISK_FILES_TOUCHED, []) or []
+      files_n = len(files) if isinstance(files, list) else 0
+      tool_calls = int(st.get(_RISK_TOOL_CALLS, 0) or 0)
+      had_shell = bool(st.get(_RISK_HAD_SHELL, False))
+      had_write = bool(st.get(_RISK_HAD_WRITE, False))
+      had_failure = bool(st.get(_RISK_HAD_FAILURE, False))
+      from gemcode.policy_profile import update_profile
+      prof = update_profile(
+        cfg.project_root,
+        files_touched=files_n,
+        tool_calls=tool_calls,
+        had_shell=had_shell,
+        had_write=had_write,
+        had_failure=had_failure,
+      )
+      object.__setattr__(cfg, "_policy_profile", prof.to_dict())
+    except Exception:
+      pass
     # ── Shell hooks: post_tool_use ────────────────────────────────────────
     try:
       from gemcode.hooks import run_post_tool_use_hook

{gemcode-0.3.69 → gemcode-0.3.71}/src/gemcode/config.py

@@ -60,6 +60,10 @@ def token_budget_invocation_reset() -> dict:
       "gemcode:bt_t0": t,
       "gemcode:bt_base_total_tokens": -1,
       "gemcode:bt_token_budget_stop": False,
+      # Tool-result aggregate budget (per user message)
+      "gemcode:tool_group_chars": 0,
+      "gemcode:tool_group_budget_exceeded": False,
+      "gemcode:tool_seq": 0,
   }
 
 
@@ -131,6 +135,12 @@ class GemCodeConfig:
     )
   )
 
+  # Aggregate tool-result budget per user message (approx. characters of tool payloads).
+  # When exceeded, GemCode tightens subsequent tool output caps for the remainder of the turn.
+  tool_result_group_budget_chars: int = field(
+    default_factory=lambda: int(os.environ.get("GEMCODE_TOOL_RESULT_GROUP_BUDGET_CHARS", "60000"))
+  )
+
   # When enabled, oversized tool outputs are offloaded to disk under
   # .gemcode/tool-results/ and replaced in history with stable refs + previews.
   # This reduces context bloat and improves prompt-cache stability.

gemcode-0.3.71/src/gemcode/policy_profile.py

@@ -0,0 +1,135 @@
+"""
+Persistent per-repo policy profile.
+
+Goal: make dynamic budgets self-tuning per repository without requiring manual
+configuration. This stores lightweight rolling stats under `.gemcode/policy.json`.
+"""
+
+from __future__ import annotations
+
+import json
+import time
+from dataclasses import dataclass
+from pathlib import Path
+from typing import Any
+
+
+def _path(root: Path) -> Path:
+  d = root / ".gemcode"
+  d.mkdir(parents=True, exist_ok=True)
+  return d / "policy.json"
+
+
+def _clamp(x: float, lo: float, hi: float) -> float:
+  return lo if x < lo else hi if x > hi else x
+
+
+def _ema(prev: float, x: float, *, alpha: float) -> float:
+  return (alpha * x) + ((1.0 - alpha) * prev)
+
+
+@dataclass(frozen=True)
+class PolicyProfile:
+  # Rolling averages in [0,1] where possible.
+  failure_rate_ema: float = 0.0
+  shell_rate_ema: float = 0.0
+  write_rate_ema: float = 0.0
+  files_touched_ema: float = 0.0  # scaled 0..1 (e.g. 0.5 ~ 10 files)
+  updated_at: int = 0
+
+  def to_dict(self) -> dict[str, Any]:
+    return {
+      "failure_rate_ema": self.failure_rate_ema,
+      "shell_rate_ema": self.shell_rate_ema,
+      "write_rate_ema": self.write_rate_ema,
+      "files_touched_ema": self.files_touched_ema,
+      "updated_at": self.updated_at,
+      "version": 1,
+    }
+
+  @staticmethod
+  def from_dict(d: dict[str, Any]) -> "PolicyProfile":
+    try:
+      return PolicyProfile(
+        failure_rate_ema=float(d.get("failure_rate_ema", 0.0) or 0.0),
+        shell_rate_ema=float(d.get("shell_rate_ema", 0.0) or 0.0),
+        write_rate_ema=float(d.get("write_rate_ema", 0.0) or 0.0),
+        files_touched_ema=float(d.get("files_touched_ema", 0.0) or 0.0),
+        updated_at=int(d.get("updated_at", 0) or 0),
+      )
+    except Exception:
+      return PolicyProfile()
+
+
+def load_profile(project_root: Path) -> PolicyProfile:
+  p = _path(project_root)
+  if not p.exists():
+    return PolicyProfile()
+  try:
+    raw = p.read_text(encoding="utf-8", errors="replace")
+    d = json.loads(raw) if raw.strip() else {}
+    if isinstance(d, dict):
+      return PolicyProfile.from_dict(d)
+  except Exception:
+    return PolicyProfile()
+  return PolicyProfile()
+
+
+def save_profile(project_root: Path, profile: PolicyProfile) -> None:
+  p = _path(project_root)
+  p.write_text(
+    json.dumps(profile.to_dict(), ensure_ascii=False, indent=2),
+    encoding="utf-8",
+    errors="replace",
+  )
+
+
+def update_profile(
+  project_root: Path,
+  *,
+  files_touched: int,
+  tool_calls: int,
+  had_shell: bool,
+  had_write: bool,
+  had_failure: bool,
+  alpha: float = 0.08,
+) -> PolicyProfile:
+  """
+  Update profile with a single-turn observation.
+
+  We scale files_touched into [0,1] via min(files/20, 1).
+  """
+  prof = load_profile(project_root)
+  alpha = _clamp(alpha, 0.01, 0.3)
+  ft_scaled = _clamp(float(files_touched) / 20.0, 0.0, 1.0)
+  fail = 1.0 if had_failure else 0.0
+  shell = 1.0 if had_shell else 0.0
+  write = 1.0 if had_write else 0.0
+  # tool_calls unused for now, but reserved for future calibration.
+  _ = tool_calls
+  updated = PolicyProfile(
+    failure_rate_ema=_ema(prof.failure_rate_ema, fail, alpha=alpha),
+    shell_rate_ema=_ema(prof.shell_rate_ema, shell, alpha=alpha),
+    write_rate_ema=_ema(prof.write_rate_ema, write, alpha=alpha),
+    files_touched_ema=_ema(prof.files_touched_ema, ft_scaled, alpha=alpha),
+    updated_at=int(time.time()),
+  )
+  save_profile(project_root, updated)
+  return updated
+
+
+def calibrated_baseline_risk(profile: PolicyProfile) -> float:
+  """
+  Convert profile into a baseline risk prior for a repo.
+
+  Repos with frequent failures, many writes, and lots of files touched tend to
+  benefit from higher evidence budgets by default.
+  """
+  r = (
+    0.55 * profile.failure_rate_ema
+    + 0.20 * profile.write_rate_ema
+    + 0.15 * profile.shell_rate_ema
+    + 0.10 * profile.files_touched_ema
+  )
+  return _clamp(r, 0.0, 0.8)
+

{gemcode-0.3.69 → gemcode-0.3.71}/src/gemcode/repl_slash.py

@@ -395,6 +395,7 @@ async def process_repl_slash(
     try:
       risk = float(getattr(cfg, "_risk_score", 0.0) or 0.0)
       pct = getattr(cfg, "_context_percent_left", None)
+      prof = getattr(cfg, "_policy_profile", None)
       out("  Dynamic policy:")
       out(f"    dynamic_token_policy:  {getattr(cfg, 'dynamic_token_policy', True)}")
       out(f"    dynamic_risk_policy:   {getattr(cfg, 'dynamic_risk_policy', True)}")
@@ -402,6 +403,12 @@ async def process_repl_slash(
       out(f"    risk_score:            {risk:.2f}")
       if isinstance(pct, int):
         out(f"    context_percent_left:  {pct}%")
+      if isinstance(prof, dict):
+        try:
+          out(f"    profile.failure_rate_ema: {float(prof.get('failure_rate_ema', 0.0) or 0.0):.2f}")
+          out(f"    profile.files_touched_ema: {float(prof.get('files_touched_ema', 0.0) or 0.0):.2f}")
+        except Exception:
+          pass
       out()
     except Exception:
       pass

{gemcode-0.3.69 → gemcode-0.3.71}/src/gemcode/session_runtime.py

@@ -314,6 +314,16 @@ def _build_artifact_service(cfg: GemCodeConfig):
 
 def create_runner(cfg: GemCodeConfig, extra_tools: list | None = None) -> Runner:
   """Construct Runner + SQLite session service + root LlmAgent."""
+  # Load per-repo calibration profile (self-tuning dynamic policy).
+  try:
+    from gemcode.policy_profile import calibrated_baseline_risk, load_profile
+    prof = load_profile(cfg.project_root)
+    base = calibrated_baseline_risk(prof)
+    cur = float(getattr(cfg, "_risk_score", 0.0) or 0.0)
+    object.__setattr__(cfg, "_risk_score", max(cur, base))
+    object.__setattr__(cfg, "_policy_profile", prof.to_dict())
+  except Exception:
+    pass
   modality_tools = build_modality_extra_tools(cfg)
   merged_extra_tools: list | None
   if extra_tools:

{gemcode-0.3.69 → gemcode-0.3.71}/src/gemcode/tool_result_store.py

@@ -18,6 +18,73 @@ from pathlib import Path
 from typing import Any
 
 _REF_PREFIX = "tool_result:"
+_REPL_STATE_KEY = "gemcode:tool_replacement_state"
+
+
+def _stable_key(tool_name: str, args: dict[str, Any] | None, seq: int | None) -> str:
+  """
+  Build a stable key for replacement decisions.
+
+  ADK does not expose tool_use_id directly to callbacks in all versions, so we use:
+  - per-turn tool sequence number (preferred when available)
+  - tool name
+  - a stable hash of args (best-effort)
+  """
+  import json
+  try:
+    args_s = json.dumps(args or {}, sort_keys=True, ensure_ascii=False)
+  except Exception:
+    args_s = str(args or {})
+  b = (f"{seq or 0}:{tool_name}:{args_s}").encode("utf-8", errors="replace")
+  return _sha256_bytes(b)
+
+
+def maybe_offload_tool_result_stable(
+  *,
+  project_root: Path,
+  tool_name: str,
+  args: dict[str, Any] | None,
+  payload: Any,
+  max_inline_chars: int,
+  state: dict[str, Any] | None,
+  seq: int | None,
+) -> tuple[Any, bool]:
+  """
+  Stable offload wrapper.
+
+  - If we've already processed an identical tool call in this session, we re-apply
+    the exact same replacement structure to preserve prompt byte stability.
+  - Otherwise, we apply offload and remember the replacement result.
+  """
+  if state is None:
+    return maybe_offload_tool_result(
+      project_root=project_root,
+      tool_name=tool_name,
+      payload=payload,
+      max_inline_chars=max_inline_chars,
+    )
+
+  repl_state = state.get(_REPL_STATE_KEY)
+  if not isinstance(repl_state, dict):
+    repl_state = {}
+    state[_REPL_STATE_KEY] = repl_state
+
+  key = _stable_key(tool_name, args, seq)
+  if key in repl_state:
+    return repl_state[key], False
+
+  new_payload, did = maybe_offload_tool_result(
+    project_root=project_root,
+    tool_name=tool_name,
+    payload=payload,
+    max_inline_chars=max_inline_chars,
+  )
+  if did:
+    repl_state[key] = new_payload
+  else:
+    # Freeze "no replacement" decision too (prevents later shape drift).
+    repl_state[key] = payload
+  return new_payload, did
 
 
 def _store_dir(project_root: Path) -> Path:

{gemcode-0.3.69 → gemcode-0.3.71/src/gemcode.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: gemcode
-Version: 0.3.69
+Version: 0.3.71
 Summary: Local-first coding agent on Google Gemini + ADK
 Author: GemCode Contributors
 License:                                  Apache License
@@ -217,6 +217,16 @@ gemcode --yes "Add a module docstring to src/foo.py"
 gemcode --session mysess --yes "Continue: run tests and fix failures"
 ```
 
+### What GemCode writes to `.gemcode/`
+
+GemCode keeps project-local state under `.gemcode/`:
+
+- **`sessions.sqlite`**: session events/history (ADK `SqliteSessionService`)
+- **`audit.log`**: JSONL audit trail for tool usage + model usage + stop reasons
+- **`tool-results/`**: oversized tool outputs offloaded to stable refs (`tool_result:<sha>`)
+- **`artifacts/`**: file artifacts (ADK `FileArtifactService`)
+- **`policy.json`**: self-tuning per-repo profile used to calibrate dynamic budgets
+
 - **`--yes`**: allow mutating tools (`write_file`, `search_replace`). Shell execution is still restricted by the `.env.example` allowlist.
 - **`--session`**: Conversation history is stored under `.gemcode/sessions.sqlite` (ADK `SqliteSessionService`). Reuse the same `--session` id to continue.
 - **`--max-llm-calls`**: cap model↔tool iterations for this message (maps to ADK `RunConfig.max_llm_calls`). You can also set `GEMCODE_MAX_LLM_CALLS`.
@@ -243,6 +253,33 @@ gemcode --session mysess --yes "Continue: run tests and fix failures"
 - **Recovery-loop**: ADK `ReflectAndRetryToolPlugin`-based retries on tool failures.
   - Set `GEMCODE_ENABLE_TOOL_RECOVERY_RETRY=0` to disable.
   - Set `GEMCODE_TOOL_REFLECT_MAX_RETRIES=1` to control retries per tool.
+
+### Token efficiency (dynamic + intelligent)
+
+GemCode optimizes tokens without losing capability by using a dynamic policy:
+
+- **Context-pressure aware**: tool caps tighten when context is tight, loosen when there is room.
+- **Risk/complexity aware**: caps increase for risky tasks (writes, shell, failures, many files).
+- **Self-tuning per repo**: `.gemcode/policy.json` calibrates baseline evidence budgets over time.
+
+Key env toggles:
+
+- `GEMCODE_DYNAMIC_TOKEN_POLICY=0|1`
+- `GEMCODE_DYNAMIC_RISK_POLICY=0|1`
+- `GEMCODE_DYNAMIC_RISK_BOOST=<float>` (default `0.6`)
+- `GEMCODE_TOOL_RESULT_OFFLOAD=0|1` (default `1`)
+
+Live telemetry:
+
+- `/status` shows `risk_score`, `context_percent_left`, and profile EMAs.
+
+### Stable tool output offloading (OpenClaude-style)
+
+Oversized tool outputs are automatically offloaded and replaced with stable refs:
+
+- Stored in: `.gemcode/tool-results/`
+- References: `tool_result:<sha256>`
+- Load on demand: `load_tool_result(ref)`
 - **Gemini thinking controls (Claude-like)**:
   - By default GemCode lets Gemini use its dynamic/adaptive thinking behavior.
   - Set `GEMCODE_DISABLE_THINKING=1` to force a best-effort “low thinking” mode:
@@ -301,12 +338,24 @@ the user’s project.
   - `list_directory`
   - `glob_files`
   - `grep_content`
+  - `repo_map`
+  - `web_search`
+  - `web_fetch`
+  - `notebook_read`
+  - `notebook_edit`
 - Mutating tools (require `--yes` unless your policy blocks them):
   - `write_file`
   - `search_replace`
 - Shell execution:
   - `run_command` (guarded by `GEMCODE_ALLOW_COMMANDS` from `.env.example` and
     `GEMCODE_PERMISSION_MODE`).
+  - `bash` (pipelines + redirects; supports `background=True`)
+
+- Background task management (for processes started via `bash(..., background=True)`):
+  - `list_tasks`, `task_output`, `kill_task`
+
+- Tool offload loader:
+  - `load_tool_result(ref)`
 
 Tool execution is still controlled by permission gates and then governed by
 GemCode’s circuit breaker + recovery behavior.
@@ -434,6 +483,23 @@ pip install -e ".[dev]"
 pytest
 ```
 
+## Release workflow (GitHub Actions → PyPI)
+
+This repository publishes to PyPI on tag pushes:
+
+- `.github/workflows/publish-pypi.yml` triggers on tags matching `v*`
+
+Typical release steps:
+
+```bash
+# 1) bump gemcode/pyproject.toml version
+git add -A
+git commit -m "release: vX.Y.Z"
+git tag -a vX.Y.Z -m "vX.Y.Z"
+git push origin HEAD
+git push origin vX.Y.Z
+```
+
 ## References (local only)
 
 Do not commit proprietary leaked trees into this package. Keep `claude-code-leaked/` and similar folders outside version control or in a private mirror.

{gemcode-0.3.69 → gemcode-0.3.71}/src/gemcode.egg-info/SOURCES.txt

@@ -32,6 +32,7 @@ src/gemcode/model_routing.py
 src/gemcode/openapi_loader.py
 src/gemcode/paths.py
 src/gemcode/permissions.py
+src/gemcode/policy_profile.py
 src/gemcode/pricing.py
 src/gemcode/prompt_suggestions.py
 src/gemcode/refine.py