ayder-cli 2.2.9__tar.gz → 2.3.1__tar.gz

{ayder_cli-2.2.9 → ayder_cli-2.3.1}/.gitignore

@@ -24,7 +24,6 @@ htmlcov/
 .venv
 TODO
 .env
-uv.lock
 
 # Project Spesific
 .ayder

{ayder_cli-2.2.9 → ayder_cli-2.3.1}/PKG-INFO

@@ -1,6 +1,6 @@
-Metadata-Version: 2.4
+Metadata-Version: 2.5
 Name: ayder-cli
-Version: 2.2.9
+Version: 2.3.1
 Summary: AI agent for any LLMs
 Project-URL: Homepage, https://github.com/ayder/ayder-cli
 Project-URL: Repository, https://github.com/ayder/ayder-cli.git
@@ -15,7 +15,8 @@ Classifier: License :: OSI Approved :: MIT License
 Classifier: Programming Language :: Python :: 3
 Requires-Python: >=3.12
 Requires-Dist: httpx
-Requires-Dist: loguru
+Requires-Dist: loguru<0.8,>=0.7
+Requires-Dist: mcp<2,>=1.19
 Requires-Dist: ollama>=0.6.2
 Requires-Dist: openai
 Requires-Dist: python-dotenv>=1.0.0
@@ -26,17 +27,16 @@ Provides-Extra: all
 Requires-Dist: anthropic; extra == 'all'
 Requires-Dist: dashscope; extra == 'all'
 Requires-Dist: google-genai; extra == 'all'
-Requires-Dist: zhipuai; extra == 'all'
+Requires-Dist: zhipuai>=2; extra == 'all'
 Provides-Extra: anthropic
 Requires-Dist: anthropic; extra == 'anthropic'
 Provides-Extra: glm
-Requires-Dist: zhipuai; extra == 'glm'
+Requires-Dist: zhipuai>=2; extra == 'glm'
 Provides-Extra: google
 Requires-Dist: google-genai; extra == 'google'
 Provides-Extra: libcst
 Requires-Dist: libcst>=1.0.0; extra == 'libcst'
 Provides-Extra: mcp
-Requires-Dist: mcp; extra == 'mcp'
 Provides-Extra: qwen
 Requires-Dist: dashscope; extra == 'qwen'
 Description-Content-Type: text/markdown
@@ -209,10 +209,23 @@ tool_tags = ["core", "metadata"]  # Enabled tool tags (see /plugin)
 agent_timeout = 300           # Seconds before a background agent is cancelled
 
 [logging]
-file_enabled = true
-file_path = ".ayder/log/ayder.log"
-rotation = "10 MB"
-retention = "7 days"
+file_enabled  = true                     # master switch for ALL file logging
+file_path     = ".ayder/log/ayder.log"
+error_path    = ".ayder/log/errors.log"  # always written unless file_enabled = false
+trace_enabled = false
+trace_path    = ".ayder/log/trace.jsonl"
+rotation      = "10 MB"
+retention     = "7 days"
+# level = "INFO"   # optional; omitted entirely when unset, which means NONE.
+                   # NONE, ERROR, WARNING, INFO, DEBUG, TRACE
+
+[logging.channels]
+# Per-channel level overrides. Unlisted channels inherit the global level.
+# Channels: llm, tool, agent, context, plugin, ui, core, external
+#   llm      = "TRACE"     # verbose LLM traffic
+#   plugin   = "WARNING"   # quiet plugin discovery
+#   external = "ERROR"     # throttle third-party libraries
+#   ui       = "NONE"      # disable this channel entirely
 
 [context_manager]
 enabled = true
@@ -253,6 +266,50 @@ num_ctx = 1000000
 
 Please adjust `num_ctx` context size window according to your local computer RAM. If Ollama crashes, decrease the value.
 
+### Reasoning Effort (`/effort`)
+
+Use `/effort` to open a picker, or `/effort high` to change the active session.
+Changes apply between turns and are inherited by future agents that use the
+session's profile; explicit agent settings take precedence. They do not change
+`config.toml`. `/effort default` restores the provider's default behavior
+(omits OpenAI effort and Ollama's `think` parameter).
+
+| Driver | Runtime values |
+|--------|----------------|
+| `openai` | `default`, `none`, `minimal`, `low`, `medium`, `high`, `xhigh`, `max` |
+| `ollama` | `default`, `on`, `off`, `low`, `medium`, `high` (`true`/`false` and `none` are also accepted) |
+
+Support varies by model. OpenAI values follow its
+[Chat Completions API](https://developers.openai.com/api/reference/resources/chat/subresources/completions/methods/create).
+Ollama uses the installed Python SDK's boolean and three-level `think` options;
+its SDK does not yet accept the server API's newer `max` value. GPT-OSS expects
+`low`, `medium`, or `high`, and cannot disable thinking. See
+[Ollama thinking](https://docs.ollama.com/capabilities/thinking).
+
+Set persistent defaults per LLM profile and agent:
+
+```toml
+[llm.openai]
+driver = "openai"
+reasoning_effort = "medium"
+
+[llm.ollama]
+driver = "ollama"
+reasoning_effort = "high"
+
+[agents.reviewer]
+provider = "openai"
+reasoning_effort = "high"
+system_prompt = "Review code and report findings."
+```
+
+Omit an agent's `reasoning_effort` to inherit its profile, or set it to
+`"default"` to clear an inherited effort. On Ollama, `"none"` maps to
+`think = false`; an explicit effort overrides the legacy `think` setting.
+Existing `think = true/false/"low"/"medium"/"high"` profile and agent settings
+remain supported. OpenAI omits reasoning effort by default, preserving
+compatibility with models and endpoints that do not support it.
+
 ### Changing Models on the Fly (`/model`)
 
 You do **not** need a separate profile for every model. The profile defines the *connection* (driver, base URL, API key).
@@ -379,6 +436,7 @@ You can also manually manage context with the `context` tool and its slash-comma
 | `/help` | Show available commands and keyboard shortcuts |
 | `/provider` | Switch LLM provider (interactive selector or direct name) |
 | `/model` | List available models or switch model (e.g., `/model qwen3-coder`) |
+| `/effort` | Pick reasoning effort or set it directly (e.g., `/effort high`) |
 | `/plugin` | Toggle tool plugins by tag (e.g., venv, http, background, python, dbs) |
 | `/tools` | List currently enabled tools and descriptions |
 | `/permission` | Toggle permission levels (r/w/x/http) interactively |
@@ -389,7 +447,7 @@ You can also manually manage context with the `context` tool and its slash-comma
 | `/notes` | Browse and edit markdown notes |
 | `/skill` | Activate a domain skill from `.ayder/skills/` via the shared `skill` tool backend |
 | `/verbose` | Toggle verbose mode |
-| `/logging` | Set log level for current session (NONE, ERROR, WARNING, INFO, DEBUG) |
+| `/logging` | Set log level for current session (NONE, ERROR, WARNING, INFO, DEBUG, TRACE) |
 | `/compact` | Summarize conversation, save a context snapshot, clear, and reload |
 | `/save-context` | Summarize conversation and save it to a named context slot (no clear) |
 | `/load-context` | Load a saved context slot and restore it |
@@ -406,9 +464,20 @@ You can also manually manage context with the `context` tool and its slash-comma
 The LLM can also call `skill(action="list" | "load" | "unload")`
 automatically when a request clearly matches an available project skill.
 
-- Default: when logging is enabled, logs go to `.ayder/log/ayder.log` (not shown on screen).
+- Errors are always captured to `.ayder/log/errors.log`, even when
+  `logging_level` is `NONE`. Set `file_enabled = false` under `[logging]` to
+  turn off all file logging including errors.
+- Unhandled exceptions, SIGTERM and SIGINT are captured to
+  `.ayder/log/errors.log`. Records in flight are lost on `SIGKILL` (`kill -9`),
+  which no process can intercept.
+- Narrative logs go to `.ayder/log/ayder.log` when a level is set.
+- `--trace` writes structured events to `.ayder/log/trace.jsonl`. It is not a
+  log level.
+- `--log-channel llm,tool` limits console and narrative output to those
+  subsystems. It never filters the error log.
 - TUI `/logging` changes are session-only and do not modify `config.toml`.
-- CLI `--verbose [LEVEL]` enables stdout logging for that run.
+- CLI `--verbose` enables stdout logging for that run; it does not narrow
+  channels.
 
 ### Keyboard Shortcuts
 
@@ -646,9 +715,46 @@ RECOMMENDATIONS: Parameterize all DB queries; add Pydantic validators to all end
 
 This summary is injected into the main LLM's context as a system message so the main agent can act on it, summarize it for you, or chain it to further work.
 
+## MCP servers
+
+MCP support is built in and installed with Ayder. Add `.ayder/mcp.json` in your
+project root, then start Ayder:
+
+```json
+{
+  "mcpServers": {
+    "local-tools": {
+      "command": "python",
+      "args": ["server.py"],
+      "env": {"EXAMPLE_SETTING": "value"}
+    },
+    "remote-tools": {
+      "url": "http://localhost:8000/mcp"
+    }
+  }
+}
+```
+
+Local servers use stdio and run from the project root; remote servers use
+Streamable HTTP. Their tools automatically join the standard `core` toolset in
+the CLI, TUI, and delegated agents. No plugin installation or `/plugin` toggle
+is needed. Tool calls retain the usual permissions: `x` for stdio servers and
+`http` for HTTP servers. Configure only servers you intend Ayder to connect to:
+local server commands start during runtime initialization to discover tools.
+
+Tools keep their server-provided names unless they conflict with an existing
+tool; conflicting names receive a `server__` prefix (and a numeric suffix if
+needed). Unreachable servers are skipped so other tools remain available.
+Connections are shared by agents in the same project and closed on process
+exit. Restart Ayder after changing `mcp.json`.
+
+Existing `mcp-tool` plugin installations are automatically skipped in favor of
+built-in support; your existing `.ayder/mcp.json` continues to work. The legacy
+`ayder-cli[mcp]` installation extra is still accepted.
+
 ## Plugins
 
-Official plugins (venv, python, dbs, mcp, temporal) are maintained in the **[ayder/ayder-plugins](https://github.com/ayder/ayder-plugins)** repository. Install any plugin directly from GitHub:
+Official plugins (venv, python, dbs, temporal) are maintained in the **[ayder/ayder-plugins](https://github.com/ayder/ayder-plugins)** repository. Install any plugin directly from GitHub:
 
 ```bash
 ayder install-plugin https://github.com/ayder/ayder-plugins/tree/main/venv-tools

{ayder_cli-2.2.9 → ayder_cli-2.3.1}/README.md

@@ -166,10 +166,23 @@ tool_tags = ["core", "metadata"]  # Enabled tool tags (see /plugin)
 agent_timeout = 300           # Seconds before a background agent is cancelled
 
 [logging]
-file_enabled = true
-file_path = ".ayder/log/ayder.log"
-rotation = "10 MB"
-retention = "7 days"
+file_enabled  = true                     # master switch for ALL file logging
+file_path     = ".ayder/log/ayder.log"
+error_path    = ".ayder/log/errors.log"  # always written unless file_enabled = false
+trace_enabled = false
+trace_path    = ".ayder/log/trace.jsonl"
+rotation      = "10 MB"
+retention     = "7 days"
+# level = "INFO"   # optional; omitted entirely when unset, which means NONE.
+                   # NONE, ERROR, WARNING, INFO, DEBUG, TRACE
+
+[logging.channels]
+# Per-channel level overrides. Unlisted channels inherit the global level.
+# Channels: llm, tool, agent, context, plugin, ui, core, external
+#   llm      = "TRACE"     # verbose LLM traffic
+#   plugin   = "WARNING"   # quiet plugin discovery
+#   external = "ERROR"     # throttle third-party libraries
+#   ui       = "NONE"      # disable this channel entirely
 
 [context_manager]
 enabled = true
@@ -210,6 +223,50 @@ num_ctx = 1000000
 
 Please adjust `num_ctx` context size window according to your local computer RAM. If Ollama crashes, decrease the value.
 
+### Reasoning Effort (`/effort`)
+
+Use `/effort` to open a picker, or `/effort high` to change the active session.
+Changes apply between turns and are inherited by future agents that use the
+session's profile; explicit agent settings take precedence. They do not change
+`config.toml`. `/effort default` restores the provider's default behavior
+(omits OpenAI effort and Ollama's `think` parameter).
+
+| Driver | Runtime values |
+|--------|----------------|
+| `openai` | `default`, `none`, `minimal`, `low`, `medium`, `high`, `xhigh`, `max` |
+| `ollama` | `default`, `on`, `off`, `low`, `medium`, `high` (`true`/`false` and `none` are also accepted) |
+
+Support varies by model. OpenAI values follow its
+[Chat Completions API](https://developers.openai.com/api/reference/resources/chat/subresources/completions/methods/create).
+Ollama uses the installed Python SDK's boolean and three-level `think` options;
+its SDK does not yet accept the server API's newer `max` value. GPT-OSS expects
+`low`, `medium`, or `high`, and cannot disable thinking. See
+[Ollama thinking](https://docs.ollama.com/capabilities/thinking).
+
+Set persistent defaults per LLM profile and agent:
+
+```toml
+[llm.openai]
+driver = "openai"
+reasoning_effort = "medium"
+
+[llm.ollama]
+driver = "ollama"
+reasoning_effort = "high"
+
+[agents.reviewer]
+provider = "openai"
+reasoning_effort = "high"
+system_prompt = "Review code and report findings."
+```
+
+Omit an agent's `reasoning_effort` to inherit its profile, or set it to
+`"default"` to clear an inherited effort. On Ollama, `"none"` maps to
+`think = false`; an explicit effort overrides the legacy `think` setting.
+Existing `think = true/false/"low"/"medium"/"high"` profile and agent settings
+remain supported. OpenAI omits reasoning effort by default, preserving
+compatibility with models and endpoints that do not support it.
+
 ### Changing Models on the Fly (`/model`)
 
 You do **not** need a separate profile for every model. The profile defines the *connection* (driver, base URL, API key).
@@ -336,6 +393,7 @@ You can also manually manage context with the `context` tool and its slash-comma
 | `/help` | Show available commands and keyboard shortcuts |
 | `/provider` | Switch LLM provider (interactive selector or direct name) |
 | `/model` | List available models or switch model (e.g., `/model qwen3-coder`) |
+| `/effort` | Pick reasoning effort or set it directly (e.g., `/effort high`) |
 | `/plugin` | Toggle tool plugins by tag (e.g., venv, http, background, python, dbs) |
 | `/tools` | List currently enabled tools and descriptions |
 | `/permission` | Toggle permission levels (r/w/x/http) interactively |
@@ -346,7 +404,7 @@ You can also manually manage context with the `context` tool and its slash-comma
 | `/notes` | Browse and edit markdown notes |
 | `/skill` | Activate a domain skill from `.ayder/skills/` via the shared `skill` tool backend |
 | `/verbose` | Toggle verbose mode |
-| `/logging` | Set log level for current session (NONE, ERROR, WARNING, INFO, DEBUG) |
+| `/logging` | Set log level for current session (NONE, ERROR, WARNING, INFO, DEBUG, TRACE) |
 | `/compact` | Summarize conversation, save a context snapshot, clear, and reload |
 | `/save-context` | Summarize conversation and save it to a named context slot (no clear) |
 | `/load-context` | Load a saved context slot and restore it |
@@ -363,9 +421,20 @@ You can also manually manage context with the `context` tool and its slash-comma
 The LLM can also call `skill(action="list" | "load" | "unload")`
 automatically when a request clearly matches an available project skill.
 
-- Default: when logging is enabled, logs go to `.ayder/log/ayder.log` (not shown on screen).
+- Errors are always captured to `.ayder/log/errors.log`, even when
+  `logging_level` is `NONE`. Set `file_enabled = false` under `[logging]` to
+  turn off all file logging including errors.
+- Unhandled exceptions, SIGTERM and SIGINT are captured to
+  `.ayder/log/errors.log`. Records in flight are lost on `SIGKILL` (`kill -9`),
+  which no process can intercept.
+- Narrative logs go to `.ayder/log/ayder.log` when a level is set.
+- `--trace` writes structured events to `.ayder/log/trace.jsonl`. It is not a
+  log level.
+- `--log-channel llm,tool` limits console and narrative output to those
+  subsystems. It never filters the error log.
 - TUI `/logging` changes are session-only and do not modify `config.toml`.
-- CLI `--verbose [LEVEL]` enables stdout logging for that run.
+- CLI `--verbose` enables stdout logging for that run; it does not narrow
+  channels.
 
 ### Keyboard Shortcuts
 
@@ -603,9 +672,46 @@ RECOMMENDATIONS: Parameterize all DB queries; add Pydantic validators to all end
 
 This summary is injected into the main LLM's context as a system message so the main agent can act on it, summarize it for you, or chain it to further work.
 
+## MCP servers
+
+MCP support is built in and installed with Ayder. Add `.ayder/mcp.json` in your
+project root, then start Ayder:
+
+```json
+{
+  "mcpServers": {
+    "local-tools": {
+      "command": "python",
+      "args": ["server.py"],
+      "env": {"EXAMPLE_SETTING": "value"}
+    },
+    "remote-tools": {
+      "url": "http://localhost:8000/mcp"
+    }
+  }
+}
+```
+
+Local servers use stdio and run from the project root; remote servers use
+Streamable HTTP. Their tools automatically join the standard `core` toolset in
+the CLI, TUI, and delegated agents. No plugin installation or `/plugin` toggle
+is needed. Tool calls retain the usual permissions: `x` for stdio servers and
+`http` for HTTP servers. Configure only servers you intend Ayder to connect to:
+local server commands start during runtime initialization to discover tools.
+
+Tools keep their server-provided names unless they conflict with an existing
+tool; conflicting names receive a `server__` prefix (and a numeric suffix if
+needed). Unreachable servers are skipped so other tools remain available.
+Connections are shared by agents in the same project and closed on process
+exit. Restart Ayder after changing `mcp.json`.
+
+Existing `mcp-tool` plugin installations are automatically skipped in favor of
+built-in support; your existing `.ayder/mcp.json` continues to work. The legacy
+`ayder-cli[mcp]` installation extra is still accepted.
+
 ## Plugins
 
-Official plugins (venv, python, dbs, mcp, temporal) are maintained in the **[ayder/ayder-plugins](https://github.com/ayder/ayder-plugins)** repository. Install any plugin directly from GitHub:
+Official plugins (venv, python, dbs, temporal) are maintained in the **[ayder/ayder-plugins](https://github.com/ayder/ayder-plugins)** repository. Install any plugin directly from GitHub:
 
 ```bash
 ayder install-plugin https://github.com/ayder/ayder-plugins/tree/main/venv-tools

{ayder_cli-2.2.9 → ayder_cli-2.3.1}/docs/PROJECT_STRUCTURE.md

@@ -160,7 +160,8 @@ cli.py:main()
 | `core/context_manager_factory.py` | Registry-based factory (OCP) | `ContextManagerFactory`, `context_manager_factory` |
 | `core/cache_monitor.py` | Timing-based KV-cache hit detection | `CacheMonitor`, `CacheStatus`, `CacheSample` |
 | `console.py` | Rich console singleton | `console` |
-| `logging_config.py` | Logger setup (loguru + stdlib bridge) | `setup_logging()` |
+| `log.py` | Channel-scoped loguru facade; leaf module so `logging_config.py` and `core/config.py` can both depend on it | `get_logger()`, `LOG_LEVELS`, `CHANNELS`, `level_no()`, `flush()` |
+| `logging_config.py` | Logger setup (loguru + stdlib bridge) | `setup_logging()`, `LoggingSettings` |
 | `version.py` | Package version constant | `__version__` |
 
 ### Root-Level Application Modules

{ayder_cli-2.2.9 → ayder_cli-2.3.1}/docs/config.toml.example

@@ -48,12 +48,15 @@
 #      `export OLLAMA_API_KEY="<key>"`. The api_key config field is INERT on the
 #      native path, so a missing env var ⇒ 401 Unauthorized.
 #
-# NOTE on `think`: on the native `ollama` driver, ayder defaults thinking ON
-#       (think = true) and forwards either a boolean or Ollama's "low",
-#       "medium", "high" levels. If a selected model rejects thinking mode,
-#       ayder retries that request once with think=false. Set `think = false`
-#       in an [llm.*] profile for known non-thinking models. The `openai`
-#       driver sends no reasoning-effort parameter at all.
+# Reasoning defaults can be set in [llm.*] and overridden in [agents.*].
+# reasoning_effort = "default" uses existing think behavior on Ollama and omits
+# the effort parameter on OpenAI. Explicit effort overrides legacy `think`.
+# OpenAI: none, minimal, low, medium, high, xhigh, max (model-dependent).
+# Ollama SDK: none (= think=false), low, medium, high. To enable/disable
+# boolean thinking explicitly, the legacy think = true/false still works.
+# Use /effort in the TUI for a session-only override or interactive picker.
+# Explicit unsupported efforts report a server error; only legacy `think`
+# retains the automatic retry with think=false for non-thinking models.
 #
 # NOTE on tuning knobs — which actually take effect depends on the DRIVER. These
 #       agents run on the native `ollama` driver, which ignores most of them:
@@ -69,8 +72,8 @@
 #                               path it sizes the context-manager budget.
 #                               Default 65536.
 #       The per-agent temperature / max_output_tokens values below are largely
-#       INERT on this Ollama pipeline; per-agent `think` is applied when the
-#       agent uses the native ollama driver.
+#       INERT on this Ollama pipeline; per-agent reasoning_effort takes
+#       precedence over the selected profile.
 # =============================================================================
 
 config_version = "2.0"
@@ -117,10 +120,23 @@ theme = "ayder"
 palette = "auto"
 
 [logging]
-file_enabled = true
-file_path = ".ayder/log/ayder.log"
-rotation = "10 MB"
-retention = "7 days"
+file_enabled  = true                     # master switch for ALL file logging
+file_path     = ".ayder/log/ayder.log"
+error_path    = ".ayder/log/errors.log"  # always written unless file_enabled = false
+trace_enabled = false
+trace_path    = ".ayder/log/trace.jsonl"
+rotation      = "10 MB"
+retention     = "7 days"
+# level = "INFO"   # optional; omitted entirely when unset, which means NONE.
+                   # NONE, ERROR, WARNING, INFO, DEBUG, TRACE
+
+[logging.channels]
+# Per-channel level overrides. Unlisted channels inherit the global level.
+# Channels: llm, tool, agent, context, plugin, ui, core, external
+#   llm      = "TRACE"     # verbose LLM traffic
+#   plugin   = "WARNING"   # quiet plugin discovery
+#   external = "ERROR"     # throttle third-party libraries
+#   ui       = "NONE"      # disable this channel entirely
 
 [context_manager]
 enabled = true
@@ -154,6 +170,7 @@ driver = "ollama"
 base_url = "http://localhost:11434"  # route :cloud models through the LOCAL daemon (auth via ~/.ollama/id_ed25519, one-time `ollama signin`)
 model = "deepseek-v4-pro:cloud"    # orchestrator default; per-agent `model` overrides this
 num_ctx = 131072
+reasoning_effort = "default"
 think = true                       # default; set false for known non-thinking models
 # To hit ollama.com directly instead, swap the line above for:
 #   base_url = "https://ollama.com"   # no /v1; then `export OLLAMA_API_KEY=<key>` (the api_key field is inert on this path)
@@ -163,6 +180,7 @@ driver = "ollama"
 base_url = "http://localhost:11434"
 model = "qwen3.6:35b-a3b-coding-nvfp4"
 num_ctx = 32768                     # keep modest: local VRAM is the constraint
+reasoning_effort = "default"
 think = true                        # default; set false if the selected model rejects thinking
 
 # Budget/throughput alternative — swap any cloud agent's `provider` to this, or
@@ -185,7 +203,8 @@ think = true                        # default; set false if the selected model r
 # num_ctx = 65536            # sizes the context-manager budget (default 65536); NOT sent to the API
 # max_output_tokens = 8192   # -> max_tokens / max_completion_tokens (default 4096); keep WELL below num_ctx
 # stop_sequences = []        # -> stop (default [])
-# # temperature / think are NOT applied by the openai driver today (no-ops) — omitted on purpose.
+# reasoning_effort = "medium" # choose a value supported by the selected model
+# # temperature / think are NOT applied by the openai driver.
 
 # =============================================================================
 # Agents
@@ -198,7 +217,7 @@ provider = "ollama_cloud"
 model = "glm-5.2:cloud"
 temperature = 0.4
 max_output_tokens = 8192
-think = "high"
+reasoning_effort = "high"
 system_prompt = """
 You are Project Manager #1 — the requirements analyst and spec author.
 
@@ -226,7 +245,7 @@ provider = "ollama_cloud"
 model = "deepseek-v4-pro:cloud"
 temperature = 0.2
 max_output_tokens = 4096
-think = "high"
+reasoning_effort = "high"
 system_prompt = """
 You are Project Manager #2 — the independent spec reviewer.
 
@@ -255,7 +274,7 @@ provider = "ollama_cloud"
 model = "deepseek-v4-pro:cloud"
 temperature = 0.3
 max_output_tokens = 8192
-think = "high"
+reasoning_effort = "high"
 num_ctx = 262144                   # long window: hold the whole spec while splitting it into tasks (delays compaction)
 system_prompt = """
 You are the Architect — you turn an APPROVED spec into an implementation plan.
@@ -283,7 +302,7 @@ provider = "ollama_cloud"
 model = "kimi-k2.7-code:cloud"
 temperature = 0.2
 max_output_tokens = 8192
-think = "high"
+reasoning_effort = "high"
 system_prompt = """
 You are the Architecture Reviewer — independent of the Architect.
 
@@ -385,7 +404,7 @@ provider = "ollama_cloud"
 model = "kimi-k2.7-code:cloud"
 temperature = 0.1
 max_output_tokens = 4096
-think = "high"
+reasoning_effort = "high"
 system_prompt = """
 You are the Code Reviewer. You review a diff for correctness and quality.
 
@@ -411,7 +430,7 @@ provider = "ollama_cloud"
 model = "minimax-m3:cloud"
 temperature = 0.1
 max_output_tokens = 4096
-think = "high"
+reasoning_effort = "high"
 num_ctx = 262144                   # long window: holds spec + all criteria + diff + test output at once
 system_prompt = """
 You are the Acceptance Gate — the final, independent check before "done".

{ayder_cli-2.2.9 → ayder_cli-2.3.1}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "ayder-cli"
-version = "2.2.9"
+version = "2.3.1"
 description = "AI agent for any LLMs"
 readme = "README.md"
 requires-python = ">=3.12"
@@ -22,8 +22,11 @@ classifiers = [
 dependencies = [
     "openai",
     "ollama>=0.6.2",
-    "loguru",
+    # Bounded deliberately: the masking sinks wrap Loguru's private FileSink /
+    # Message / StreamSink shapes (§F5-R6), guarded by `_shape_guard()`.
+    "loguru>=0.7,<0.8",
     "httpx",
+    "mcp>=1.19,<2",
     "rich>=13.0.0",
     # Pinned exactly to match the runtime: 8.2.5+ replaced the single
     # "textual-ansi" builtin with per-theme `ansi` flags and ansi-dark/
@@ -37,9 +40,12 @@ dependencies = [
 anthropic = ["anthropic"]
 google = ["google-genai"]
 qwen = ["dashscope"]
-glm = ["zhipuai"]
+# The provider uses the v2 client; do not let MCP's shared PyJWT dependency
+# make the resolver select the incompatible v1 SDK.
+glm = ["zhipuai>=2"]
 all = ["ayder-cli[anthropic,google,qwen,glm]"]
-mcp = ["mcp"]
+# Compatibility extra: MCP is now included in every installation.
+mcp = []
 # Optional: required only by plugins that do CST-based code transforms.
 libcst = ["libcst>=1.0.0"]
 
@@ -69,7 +75,7 @@ dev = [
     "anthropic",
     "google-genai",
     "dashscope",
-    "zhipuai",
+    "zhipuai>=2",
 ]
 
 [tool.poe.tasks]
@@ -85,7 +91,16 @@ lint = "ruff check src/"
 lint-fix = "ruff check src/ --fix"
 typecheck = "mypy src/"
 format = "ruff format src/"
-check-all = ["lint", "typecheck", "test"]
+check-all = ["lint", "typecheck", "gates", "test"]
+
+[tool.poe.tasks.gates]
+help = "Run the logging AST gates"
+sequence = [
+    { cmd = "python scripts/find_percent_logging.py" },
+    { cmd = "python scripts/check_logging_imports.py" },
+    { cmd = "python scripts/check_exception_handlers.py" },
+    { cmd = "python scripts/check_log_arguments.py" },
+]
 
 [tool.poe.tasks.clean]
 help = "Deep clean cache and build artifacts"
@@ -113,3 +128,11 @@ ignore_missing_imports = false
 [[tool.mypy.overrides]]
 module = ["zhipuai", "dashscope"]
 ignore_missing_imports = true
+
+[tool.ruff.lint]
+# Explicit rule set: ruff 0.16 widened its defaults from 59 to 413 rules,
+# so an implicit default + extend-select drifts with the installed version.
+select = ["E4", "E7", "E9", "F", "TRY400", "TRY401", "BLE001", "LOG", "G"]
+
+[tool.ruff.lint.per-file-ignores]
+"tests/**/*.py" = ["BLE001", "G004"]