oy-cli 0.3.3__tar.gz → 0.3.5__tar.gz

{oy_cli-0.3.3 → oy_cli-0.3.5}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: oy-cli
-Version: 0.3.3
+Version: 0.3.5
 Summary: Tiny local coding CLI with a small tool surface
 Author: oy-cli contributors
 License-Expression: Apache-2.0
@@ -32,6 +32,8 @@ uv tool install oy-cli
 oy "add docstrings to public functions"
 ```
 
+Note: `oy` also supports a Copilot-compatible shim via `OY_SHIM=copilot` or `copilot:model` selection where available.
+
 ## Examples
 
 ```bash
@@ -63,7 +65,9 @@ oy --help                 # Show all commands
 
 Most AI coding tools are large, complex, or lock you into a single provider. `oy` is deliberately small, easy to audit, and built around a narrow tool surface.
 
-**Design goals:** small auditable codebase, minimal tool surface, OpenAI-completions-focused CLI loop, multiple backends behind shims, fresh session each run, and explicit checkpoints when needed.
+**Design goals:** small auditable codebase, minimal tool surface,
+OpenAI-completions-focused CLI loop, multiple backends behind shims,
+fresh session each run, and explicit checkpoints when needed.
 
 ## System Prompts
 
@@ -114,7 +118,19 @@ Each tool description is passed directly to the model. These are the exact descr
 ### Audit Prompt
 
 ```markdown
-Audit the repo for security, unnecessary complexity, and major obvious performance issues, preserving project and human context. First read key markdown docs, then refresh or generate an audit header at the top of ISSUES.md that includes the current date, the latest Git commit reference, and a concise codebase summary using tools like `scc` or `tokei`. Next, fetch the current OWASP ASVS (or MASVS if more relevant) and grugbrain.dev guidelines using httpx, inspect the codebase against these, and write or merge prioritised findings (max 10-15) into the ISSUES.md file. Ensure each finding is formatted to include its location, category (security, complexity, or performance), standard reference, a clear recommendation, and has a Status, if existing findings have been resolved, summarise and note them in a short log at the end.
+Audit the repo for security, unnecessary complexity, and major
+obvious performance issues, preserving project and human context.
+First read key markdown docs, then refresh or generate an audit
+header at the top of ISSUES.md that includes the current date,
+the latest Git commit reference, and a concise codebase summary
+using tools like `scc` or `tokei`. Next, fetch the current OWASP
+ASVS (or MASVS if more relevant) and grugbrain.dev guidelines
+using httpx, inspect the codebase against these, and write or
+merge prioritised findings (max 10-15) into the ISSUES.md file.
+Ensure each finding is formatted to include its location, category
+(security, complexity, or performance), standard reference, a clear
+recommendation, and has a Status, if existing findings have been
+resolved, summarise and note them in a short log at the end.
 ```
 
 ```bash
@@ -141,18 +157,27 @@ OY_ROOT=./src oy audit      # Audit specific directory
 {"shim": "openai", "model": "glm-5"}
 ```
 
-The `shim` field pins which backend to use regardless of what else is signed in. Use `oy model <filter>` to pick interactively; it merges models from available signed-in shims into a single list using `shim:model` prefixes.
+The `shim` field pins which backend to use regardless of what else is signed in.
+Use `oy model <filter>` to pick interactively; it merges models from available
+signed-in shims into a single list using `shim:model` prefixes.
 
-On first run, if no model is configured, `oy` prompts you to pick one from the available backends. Set `OY_MODEL`, `OY_SHIM`, or save a config with `oy model` to pin behavior.
+On first run, if no model is configured, `oy` prompts you to pick one from
+the available backends. Set `OY_MODEL`, `OY_SHIM`, or save a config with
+`oy model` to pin behavior.
 
-**Recommended model:** From testing, `glm-5` offers the best balance of intelligence, cost, and tool-use ability. `kimi-k2.5` is another strong option. The [Artificial Analysis Comparison of Open Source Models](https://artificialanalysis.ai/models/open-source) is a good reference.
+**Recommended model:** From testing, `glm-5` offers the best balance of
+intelligence, cost, and tool-use ability. `kimi-k2.5` is another strong option.
+The [Artificial Analysis Comparison of Open Source Models](https://artificialanalysis.ai/models/open-source)
+is a good reference.
 
 ## Requirements
 
 - Python 3.14+
 - `bash`
 - (Optional) `rg` (ripgrep) for faster search
-- OpenAI API key or Codex local auth **OR** Gemini CLI OAuth credentials (`~/.gemini/oauth_creds.json`) **OR** Claude Code local auth **OR** AWS CLI configured for Bedrock
+- OpenAI API key or Codex local auth **OR** Gemini CLI OAuth credentials
+  (`~/.gemini/oauth_creds.json`) **OR** Claude Code local auth **OR**
+  AWS CLI configured for Bedrock
 
 ## Installation
 
@@ -174,7 +199,8 @@ export OPENAI_BASE_URL=https://your-endpoint.example/v1
 export OPENAI_API_KEY=...
 ```
 
-Gemini, Claude, Codex (OpenAI) creds should be automatically introspected and used, if creds are available `oy model` will show them in the model list.
+Gemini, Claude, Codex (OpenAI) creds should be automatically introspected
+and used, if creds are available `oy model` will show them in the model list.
 
 **AWS Bedrock (automatic):** Uses your default AWS profile/region. Supports auto-refresh of stale SSO sessions.
 ```bash
@@ -184,7 +210,9 @@ export AWS_REGION=us-west-2
 
 ## Troubleshooting
 
-**"Missing API credentials"** -> Set `OPENAI_API_KEY`, sign in with `codex`, `gemini` or `claude`, or configure AWS CLI (`aws configure`). For Bedrock: ensure your profile has `bedrock:InvokeModel` permission.
+**"Missing API credentials"** -> Set `OPENAI_API_KEY`, sign in with `codex`,
+`gemini` or `claude`, or configure AWS CLI (`aws configure`). For Bedrock:
+ensure your profile has `bedrock:InvokeModel` permission.
 
 **"stdin is not a TTY"** -> Piping input disables `ask`. Set `OY_NON_INTERACTIVE=1` to make explicit.
 
@@ -200,7 +228,9 @@ Recommended:
 - avoid exposing long-lived secrets in the environment
 - review generated changes before shipping
 
-**Automatic protections:** workspace-bound file access, structured edits through `apply`, sensitive header redaction in `httpx`, and native boto3 credential resolution for Bedrock.
+**Automatic protections:** workspace-bound file access, structured edits
+through `apply`, sensitive header redaction in `httpx`, and native boto3
+credential resolution for Bedrock.
 
 ## Links
 

{oy_cli-0.3.3 → oy_cli-0.3.5}/README.md

@@ -9,6 +9,8 @@ uv tool install oy-cli
 oy "add docstrings to public functions"
 ```
 
+Note: `oy` also supports a Copilot-compatible shim via `OY_SHIM=copilot` or `copilot:model` selection where available.
+
 ## Examples
 
 ```bash
@@ -40,7 +42,9 @@ oy --help                 # Show all commands
 
 Most AI coding tools are large, complex, or lock you into a single provider. `oy` is deliberately small, easy to audit, and built around a narrow tool surface.
 
-**Design goals:** small auditable codebase, minimal tool surface, OpenAI-completions-focused CLI loop, multiple backends behind shims, fresh session each run, and explicit checkpoints when needed.
+**Design goals:** small auditable codebase, minimal tool surface,
+OpenAI-completions-focused CLI loop, multiple backends behind shims,
+fresh session each run, and explicit checkpoints when needed.
 
 ## System Prompts
 
@@ -91,7 +95,19 @@ Each tool description is passed directly to the model. These are the exact descr
 ### Audit Prompt
 
 ```markdown
-Audit the repo for security, unnecessary complexity, and major obvious performance issues, preserving project and human context. First read key markdown docs, then refresh or generate an audit header at the top of ISSUES.md that includes the current date, the latest Git commit reference, and a concise codebase summary using tools like `scc` or `tokei`. Next, fetch the current OWASP ASVS (or MASVS if more relevant) and grugbrain.dev guidelines using httpx, inspect the codebase against these, and write or merge prioritised findings (max 10-15) into the ISSUES.md file. Ensure each finding is formatted to include its location, category (security, complexity, or performance), standard reference, a clear recommendation, and has a Status, if existing findings have been resolved, summarise and note them in a short log at the end.
+Audit the repo for security, unnecessary complexity, and major
+obvious performance issues, preserving project and human context.
+First read key markdown docs, then refresh or generate an audit
+header at the top of ISSUES.md that includes the current date,
+the latest Git commit reference, and a concise codebase summary
+using tools like `scc` or `tokei`. Next, fetch the current OWASP
+ASVS (or MASVS if more relevant) and grugbrain.dev guidelines
+using httpx, inspect the codebase against these, and write or
+merge prioritised findings (max 10-15) into the ISSUES.md file.
+Ensure each finding is formatted to include its location, category
+(security, complexity, or performance), standard reference, a clear
+recommendation, and has a Status, if existing findings have been
+resolved, summarise and note them in a short log at the end.
 ```
 
 ```bash
@@ -118,18 +134,27 @@ OY_ROOT=./src oy audit      # Audit specific directory
 {"shim": "openai", "model": "glm-5"}
 ```
 
-The `shim` field pins which backend to use regardless of what else is signed in. Use `oy model <filter>` to pick interactively; it merges models from available signed-in shims into a single list using `shim:model` prefixes.
+The `shim` field pins which backend to use regardless of what else is signed in.
+Use `oy model <filter>` to pick interactively; it merges models from available
+signed-in shims into a single list using `shim:model` prefixes.
 
-On first run, if no model is configured, `oy` prompts you to pick one from the available backends. Set `OY_MODEL`, `OY_SHIM`, or save a config with `oy model` to pin behavior.
+On first run, if no model is configured, `oy` prompts you to pick one from
+the available backends. Set `OY_MODEL`, `OY_SHIM`, or save a config with
+`oy model` to pin behavior.
 
-**Recommended model:** From testing, `glm-5` offers the best balance of intelligence, cost, and tool-use ability. `kimi-k2.5` is another strong option. The [Artificial Analysis Comparison of Open Source Models](https://artificialanalysis.ai/models/open-source) is a good reference.
+**Recommended model:** From testing, `glm-5` offers the best balance of
+intelligence, cost, and tool-use ability. `kimi-k2.5` is another strong option.
+The [Artificial Analysis Comparison of Open Source Models](https://artificialanalysis.ai/models/open-source)
+is a good reference.
 
 ## Requirements
 
 - Python 3.14+
 - `bash`
 - (Optional) `rg` (ripgrep) for faster search
-- OpenAI API key or Codex local auth **OR** Gemini CLI OAuth credentials (`~/.gemini/oauth_creds.json`) **OR** Claude Code local auth **OR** AWS CLI configured for Bedrock
+- OpenAI API key or Codex local auth **OR** Gemini CLI OAuth credentials
+  (`~/.gemini/oauth_creds.json`) **OR** Claude Code local auth **OR**
+  AWS CLI configured for Bedrock
 
 ## Installation
 
@@ -151,7 +176,8 @@ export OPENAI_BASE_URL=https://your-endpoint.example/v1
 export OPENAI_API_KEY=...
 ```
 
-Gemini, Claude, Codex (OpenAI) creds should be automatically introspected and used, if creds are available `oy model` will show them in the model list.
+Gemini, Claude, Codex (OpenAI) creds should be automatically introspected
+and used, if creds are available `oy model` will show them in the model list.
 
 **AWS Bedrock (automatic):** Uses your default AWS profile/region. Supports auto-refresh of stale SSO sessions.
 ```bash
@@ -161,7 +187,9 @@ export AWS_REGION=us-west-2
 
 ## Troubleshooting
 
-**"Missing API credentials"** -> Set `OPENAI_API_KEY`, sign in with `codex`, `gemini` or `claude`, or configure AWS CLI (`aws configure`). For Bedrock: ensure your profile has `bedrock:InvokeModel` permission.
+**"Missing API credentials"** -> Set `OPENAI_API_KEY`, sign in with `codex`,
+`gemini` or `claude`, or configure AWS CLI (`aws configure`). For Bedrock:
+ensure your profile has `bedrock:InvokeModel` permission.
 
 **"stdin is not a TTY"** -> Piping input disables `ask`. Set `OY_NON_INTERACTIVE=1` to make explicit.
 
@@ -177,7 +205,9 @@ Recommended:
 - avoid exposing long-lived secrets in the environment
 - review generated changes before shipping
 
-**Automatic protections:** workspace-bound file access, structured edits through `apply`, sensitive header redaction in `httpx`, and native boto3 credential resolution for Bedrock.
+**Automatic protections:** workspace-bound file access, structured edits
+through `apply`, sensitive header redaction in `httpx`, and native boto3
+credential resolution for Bedrock.
 
 ## Links
 

{oy_cli-0.3.3 → oy_cli-0.3.5}/oy_cli.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: oy-cli
-Version: 0.3.3
+Version: 0.3.5
 Summary: Tiny local coding CLI with a small tool surface
 Author: oy-cli contributors
 License-Expression: Apache-2.0
@@ -32,6 +32,8 @@ uv tool install oy-cli
 oy "add docstrings to public functions"
 ```
 
+Note: `oy` also supports a Copilot-compatible shim via `OY_SHIM=copilot` or `copilot:model` selection where available.
+
 ## Examples
 
 ```bash
@@ -63,7 +65,9 @@ oy --help                 # Show all commands
 
 Most AI coding tools are large, complex, or lock you into a single provider. `oy` is deliberately small, easy to audit, and built around a narrow tool surface.
 
-**Design goals:** small auditable codebase, minimal tool surface, OpenAI-completions-focused CLI loop, multiple backends behind shims, fresh session each run, and explicit checkpoints when needed.
+**Design goals:** small auditable codebase, minimal tool surface,
+OpenAI-completions-focused CLI loop, multiple backends behind shims,
+fresh session each run, and explicit checkpoints when needed.
 
 ## System Prompts
 
@@ -114,7 +118,19 @@ Each tool description is passed directly to the model. These are the exact descr
 ### Audit Prompt
 
 ```markdown
-Audit the repo for security, unnecessary complexity, and major obvious performance issues, preserving project and human context. First read key markdown docs, then refresh or generate an audit header at the top of ISSUES.md that includes the current date, the latest Git commit reference, and a concise codebase summary using tools like `scc` or `tokei`. Next, fetch the current OWASP ASVS (or MASVS if more relevant) and grugbrain.dev guidelines using httpx, inspect the codebase against these, and write or merge prioritised findings (max 10-15) into the ISSUES.md file. Ensure each finding is formatted to include its location, category (security, complexity, or performance), standard reference, a clear recommendation, and has a Status, if existing findings have been resolved, summarise and note them in a short log at the end.
+Audit the repo for security, unnecessary complexity, and major
+obvious performance issues, preserving project and human context.
+First read key markdown docs, then refresh or generate an audit
+header at the top of ISSUES.md that includes the current date,
+the latest Git commit reference, and a concise codebase summary
+using tools like `scc` or `tokei`. Next, fetch the current OWASP
+ASVS (or MASVS if more relevant) and grugbrain.dev guidelines
+using httpx, inspect the codebase against these, and write or
+merge prioritised findings (max 10-15) into the ISSUES.md file.
+Ensure each finding is formatted to include its location, category
+(security, complexity, or performance), standard reference, a clear
+recommendation, and has a Status, if existing findings have been
+resolved, summarise and note them in a short log at the end.
 ```
 
 ```bash
@@ -141,18 +157,27 @@ OY_ROOT=./src oy audit      # Audit specific directory
 {"shim": "openai", "model": "glm-5"}
 ```
 
-The `shim` field pins which backend to use regardless of what else is signed in. Use `oy model <filter>` to pick interactively; it merges models from available signed-in shims into a single list using `shim:model` prefixes.
+The `shim` field pins which backend to use regardless of what else is signed in.
+Use `oy model <filter>` to pick interactively; it merges models from available
+signed-in shims into a single list using `shim:model` prefixes.
 
-On first run, if no model is configured, `oy` prompts you to pick one from the available backends. Set `OY_MODEL`, `OY_SHIM`, or save a config with `oy model` to pin behavior.
+On first run, if no model is configured, `oy` prompts you to pick one from
+the available backends. Set `OY_MODEL`, `OY_SHIM`, or save a config with
+`oy model` to pin behavior.
 
-**Recommended model:** From testing, `glm-5` offers the best balance of intelligence, cost, and tool-use ability. `kimi-k2.5` is another strong option. The [Artificial Analysis Comparison of Open Source Models](https://artificialanalysis.ai/models/open-source) is a good reference.
+**Recommended model:** From testing, `glm-5` offers the best balance of
+intelligence, cost, and tool-use ability. `kimi-k2.5` is another strong option.
+The [Artificial Analysis Comparison of Open Source Models](https://artificialanalysis.ai/models/open-source)
+is a good reference.
 
 ## Requirements
 
 - Python 3.14+
 - `bash`
 - (Optional) `rg` (ripgrep) for faster search
-- OpenAI API key or Codex local auth **OR** Gemini CLI OAuth credentials (`~/.gemini/oauth_creds.json`) **OR** Claude Code local auth **OR** AWS CLI configured for Bedrock
+- OpenAI API key or Codex local auth **OR** Gemini CLI OAuth credentials
+  (`~/.gemini/oauth_creds.json`) **OR** Claude Code local auth **OR**
+  AWS CLI configured for Bedrock
 
 ## Installation
 
@@ -174,7 +199,8 @@ export OPENAI_BASE_URL=https://your-endpoint.example/v1
 export OPENAI_API_KEY=...
 ```
 
-Gemini, Claude, Codex (OpenAI) creds should be automatically introspected and used, if creds are available `oy model` will show them in the model list.
+Gemini, Claude, Codex (OpenAI) creds should be automatically introspected
+and used, if creds are available `oy model` will show them in the model list.
 
 **AWS Bedrock (automatic):** Uses your default AWS profile/region. Supports auto-refresh of stale SSO sessions.
 ```bash
@@ -184,7 +210,9 @@ export AWS_REGION=us-west-2
 
 ## Troubleshooting
 
-**"Missing API credentials"** -> Set `OPENAI_API_KEY`, sign in with `codex`, `gemini` or `claude`, or configure AWS CLI (`aws configure`). For Bedrock: ensure your profile has `bedrock:InvokeModel` permission.
+**"Missing API credentials"** -> Set `OPENAI_API_KEY`, sign in with `codex`,
+`gemini` or `claude`, or configure AWS CLI (`aws configure`). For Bedrock:
+ensure your profile has `bedrock:InvokeModel` permission.
 
 **"stdin is not a TTY"** -> Piping input disables `ask`. Set `OY_NON_INTERACTIVE=1` to make explicit.
 
@@ -200,7 +228,9 @@ Recommended:
 - avoid exposing long-lived secrets in the environment
 - review generated changes before shipping
 
-**Automatic protections:** workspace-bound file access, structured edits through `apply`, sensitive header redaction in `httpx`, and native boto3 credential resolution for Bedrock.
+**Automatic protections:** workspace-bound file access, structured edits
+through `apply`, sensitive header redaction in `httpx`, and native boto3
+credential resolution for Bedrock.
 
 ## Links
 

{oy_cli-0.3.3 → oy_cli-0.3.5}/oy_cli.py

@@ -4,7 +4,6 @@ from dataclasses import dataclass
 import json
 import logging
 import os
-import readline
 import re
 import sys
 import tempfile
@@ -89,8 +88,9 @@ def _init_debug_log() -> tuple[logging.Logger | None, str | None]:
     raw = os.environ.get("OY_DEBUG", "").strip().lower()
     if raw not in {"1", "true", "yes", "on"}:
         return None, None
-    fd, path = tempfile.mkstemp(prefix="oy-debug-", suffix=".jsonl")
-    os.close(fd)
+    debug_dir = CONFIG_PATH.parent
+    debug_dir.mkdir(parents=True, exist_ok=True)
+    path = str(debug_dir / "debug.jsonl")
     logger = logging.getLogger("oy.debug")
     logger.setLevel(logging.DEBUG)
     logger.propagate = False
@@ -256,7 +256,7 @@ def abort(m, c=1):
 def clip_tokens(text, limit=MAX_TOOL_OUTPUT_TOKENS, tail=0):
     """Truncate *text* to *limit* tokens, optionally keeping *tail* tokens from the end."""
     e = get_tokenizer()
-    ids = e.encode(text)
+    ids = e.encode(text, disallowed_special=())
     n = len(ids)
     if n <= limit:
         return text
@@ -312,15 +312,23 @@ def _http_body(text, ct):
     )
 
 
+_JSON_PATH_MAX_DEPTH = 20
+
+
 def _json_path(v, p):
     """Walk into *v* using dot-separated *p* (supports dict keys and list indices)."""
-    for part in (p or "").split("."):
+    for i, part in enumerate((p or "").split(".")):
         if not part:
             continue
+        if i >= _JSON_PATH_MAX_DEPTH:
+            raise ValueError(f"json_path exceeded max depth of {_JSON_PATH_MAX_DEPTH}")
         if isinstance(v, list):
             if not part.isdigit():
                 raise ValueError(f"json_path expected index, got {part}")
-            v = v[int(part)]
+            try:
+                v = v[int(part)]
+            except IndexError:
+                raise ValueError(f"json_path index {part} out of range (length {len(v)})")
         elif isinstance(v, dict):
             if part not in v:
                 raise ValueError(f"json_path key not found: {part}")
@@ -1419,21 +1427,36 @@ def tool_httpx(
     if parsed.scheme not in {"http", "https"}:
         raise ValueError("httpx only supports http and https")
     _print("status", "Fetching HTTP content.", err=True)
+    max_bytes = max_tokens * 16  # generous: ~16 bytes per token
     try:
         with httpx.Client(
-            follow_redirects=True, timeout=float(timeout_seconds)
+            follow_redirects=True, timeout=float(timeout_seconds), max_redirects=10,
         ) as http:
-            response = http.request(
+            with http.stream(
                 method,
                 parsed.geturl(),
                 headers=_norm_map(headers, "headers"),
                 params=_norm_map(params, "params"),
                 content=body,
                 json=json_body,
-            )
+            ) as response:
+                chunks: list[bytes] = []
+                total = 0
+                for chunk in response.iter_bytes():
+                    chunks.append(chunk)
+                    total += len(chunk)
+                    if total > max_bytes:
+                        break
     except httpx.HTTPError as exc:
         raise ValueError(_httpx_err(exc, timeout_seconds)) from exc
-    out = render_httpx_output(response, response_mode, json_path=json_path)
+    # Reconstruct a response with bounded content for render_httpx_output
+    bounded = httpx.Response(
+        status_code=response.status_code,
+        headers=response.headers,
+        content=b"".join(chunks)[:max_bytes],
+        request=response.request,
+    )
+    out = render_httpx_output(bounded, response_mode, json_path=json_path)
     show(out, 1)
     return clip_tokens(out, max_tokens)
 
@@ -1513,7 +1536,7 @@ def truncate_str_to_tokens(text: str, max_tokens: int = MAX_MESSAGE_TOKENS) -> s
     characters were removed so the model knows the content was cut.
     """
     enc = get_tokenizer()
-    ids = enc.encode(text)
+    ids = enc.encode(text, disallowed_special=())
     if len(ids) <= max_tokens:
         return text
     kept = enc.decode(ids[:max_tokens])
@@ -1757,6 +1780,10 @@ def audit(prompt: str = ""):
 
 def _setup_readline():
     """Configure readline with persistent history for shell-like UX."""
+    try:
+        import readline
+    except ImportError:
+        return  # no readline on minimal builds (Alpine, WASM)
     history_path = CONFIG_PATH.parent / "history"
     history_path.parent.mkdir(parents=True, exist_ok=True)
     try:
@@ -1764,24 +1791,76 @@ def _setup_readline():
     except FileNotFoundError:
         pass
     readline.set_history_length(1000)
+    # Ensure history file has restrictive permissions (M2: OWASP ASVS V8.3.4)
+    history_path.touch(mode=0o600, exist_ok=True)
     import atexit
 
     atexit.register(readline.write_history_file, str(history_path))
 
 
+def _drain_stdin(timeout: float = 0.05) -> str:
+    """Read any data already buffered on stdin (e.g. the tail of a paste).
+
+    Uses select() with a short timeout.  Returns the extra text, or "".
+    Only works on real ttys; returns "" for piped stdin.
+    """
+    import select
+    if not sys.stdin.isatty():
+        return ""
+    chunks: list[str] = []
+    while True:
+        ready, _, _ = select.select([sys.stdin], [], [], timeout)
+        if not ready:
+            break
+        chunk = os.read(sys.stdin.fileno(), 4096)
+        if not chunk:
+            break
+        chunks.append(chunk.decode("utf-8", errors="replace"))
+        # After first chunk, use a tighter timeout for the rest.
+        timeout = 0.01
+    return "".join(chunks)
+
+
 def _read_input():
-    """Read user input, supporting \\ continuation for multi-line."""
+    '''Read user input, with automatic paste detection.
+
+    Input modes:
+    1. Single line  -- type and press Enter.
+    2. Paste        -- paste multiline text; lines that arrive within a
+       few milliseconds of Enter are collected automatically.
+    3. Block mode   -- start with ``"""`` to open a fenced block;
+       close it with ``"""`` on its own line.
+
+    Paste detection works by draining stdin right after readline returns.
+    During normal typing there is nothing buffered, so it is a no-op.
+    During a paste, the remaining lines are already queued up.
+    '''
     line = input("oy > ")
-    if not line.endswith("\\"):
-        return line
-    parts = [line[:-1]]
-    while True:
-        cont = input("... ")
-        if not cont.endswith("\\"):
+
+    # --- block mode: triple-quote fence (still supported) ------------------
+    stripped = line.strip()
+    if stripped == '"""' or stripped.startswith('"""'):
+        if stripped == '"""':
+            parts: list[str] = []
+        else:
+            parts = [stripped[3:]]
+        while True:
+            try:
+                cont = input('... ')
+            except EOFError:
+                break
+            if cont.strip() == '"""':
+                break
             parts.append(cont)
-            break
-        parts.append(cont[:-1])
-    return "\n".join(parts)
+        return "\n".join(parts)
+
+    # --- paste detection: drain any remaining buffered input ---------------
+    extra = _drain_stdin()
+    if extra:
+        # Strip trailing newline that the terminal added from the final Enter.
+        return line + "\n" + extra.rstrip("\n")
+
+    return line
 
 
 
@@ -1850,7 +1929,8 @@ def _chat_command(cmd, transcript, system_prompt, model_spec):
             "- `/clear` -- reset conversation (keeps system prompt)",
             "- `/quit` or `/exit` -- end session",
             "",
-            "Tip: end a line with `\\` to continue on the next line.",
+            "Tip: paste multiline text — extra lines are detected automatically.",
+            'Tip: type `"""` to start a multiline block, `"""` to end it.',
         ]), err=True)
         return True
     if cmd == "/tokens":

{oy_cli-0.3.3 → oy_cli-0.3.5}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "oy-cli"
-version = "0.3.3"
+version = "0.3.5"
 description = "Tiny local coding CLI with a small tool surface"
 readme = "README.md"
 requires-python = ">=3.14"

{oy_cli-0.3.3 → oy_cli-0.3.5}/shim.py

@@ -13,6 +13,7 @@ from dataclasses import dataclass
 from datetime import datetime, timezone
 from email.utils import parsedate_to_datetime
 from functools import lru_cache
+from types import MappingProxyType
 from pathlib import Path
 from typing import Any, Awaitable, Callable, TypeAlias
 from urllib.parse import quote
@@ -35,6 +36,7 @@ SHIM_GEMINI = "gemini"
 SHIM_BEDROCK = "bedrock"
 SHIM_MANTLE = "bedrock-mantle"
 SHIM_CLAUDE = "claude"
+SHIM_COPILOT = "copilot"
 SHIM_ORDER = (
     SHIM_OPENAI,
     SHIM_CODEX,
@@ -42,6 +44,7 @@ SHIM_ORDER = (
     SHIM_BEDROCK,
     SHIM_MANTLE,
     SHIM_CLAUDE,
+    SHIM_COPILOT,
 )
 KNOWN_SHIMS = set(SHIM_ORDER)
 
@@ -567,7 +570,11 @@ def run_cmd(cmd, cwd=None, env=None, timeout=120, stdin_text=None):
 # If env vars change mid-process (e.g. in tests), the cache will be stale.
 @lru_cache(maxsize=8)
 def command_env(cwd=None):
-    """Build a shell environment dict, merging Homebrew and mise paths."""
+    """Build a shell environment dict, merging Homebrew and mise paths.
+
+    Returns a read-only ``MappingProxyType`` so the cached value cannot be
+    accidentally mutated by callers.
+    """
     env = os.environ.copy()
     if brew := which("brew", env.get("PATH")):
         prefix = Path(brew).parent.parent
@@ -576,7 +583,7 @@ def command_env(cwd=None):
             [str(prefix / "bin"), str(prefix / "sbin")], split_path(env.get("PATH"))
         )
     if not (mise := which("mise", env.get("PATH"))):
-        return env
+        return MappingProxyType(env)
     try:
         result = run_cmd(
             [mise, "env", "--json"],
@@ -598,7 +605,7 @@ def command_env(cwd=None):
             if key == "PATH"
             else value
         )
-    return merged
+    return MappingProxyType(merged)
 
 
 def http_client(**kw):
@@ -1640,9 +1647,6 @@ def _decode_responses_output(response: Any) -> AssistantMessage:
     )
 
 
-parse_tool_call_arguments = _decode_tool_call_arguments
-_responses_output_to_message = _decode_responses_output
-
 
 def _http_error_message(prefix: str, response: httpx.Response) -> str:
     try:
@@ -2484,6 +2488,160 @@ def _list_mantle_models(
     return _build_mantle_client(region, cwd).list_models()
 
 
+# ---------------------------------------------------------------------------
+# Copilot shim – uses the GitHub Copilot API (api.githubcopilot.com)
+# with a GitHub PAT obtained from COPILOT_GITHUB_TOKEN / GH_TOKEN /
+# GITHUB_TOKEN / `gh auth token`.  Set COPILOT_BASE_URL to override
+# (e.g. https://api.business.githubcopilot.com for enterprise).
+# Models that support /responses use that API; others fall back to
+# /chat/completions automatically.
+# ---------------------------------------------------------------------------
+
+_COPILOT_BASE_URL = os.environ.get(
+    "COPILOT_BASE_URL", "https://api.githubcopilot.com"
+)
+_COPILOT_INTEGRATION_ID = "copilot-developer-cli"
+_COPILOT_EDITOR_VERSION = "copilot-developer-cli/1.0.6"
+
+
+def _get_github_token() -> str | None:
+    """Return a GitHub PAT from env vars or the ``gh`` CLI, or *None*.
+
+    Checks (in order): ``COPILOT_GITHUB_TOKEN``, ``GH_TOKEN``,
+    ``GITHUB_TOKEN``, then ``gh auth token``.
+    """
+    for var in ("COPILOT_GITHUB_TOKEN", "GH_TOKEN", "GITHUB_TOKEN"):
+        val = os.environ.get(var)
+        if isinstance(val, str) and val:
+            return val
+    gh = which("gh")
+    if not gh:
+        return None
+    try:
+        proc = subprocess.run(
+            [gh, "auth", "token"],
+            capture_output=True,
+            text=True,
+            timeout=10,
+        )
+        token = proc.stdout.strip()
+        return token if proc.returncode == 0 and token else None
+    except Exception:
+        return None
+
+
+def _copilot_default_headers() -> dict[str, str]:
+    return {
+        "Copilot-Integration-Id": _COPILOT_INTEGRATION_ID,
+        "Editor-Version": _COPILOT_EDITOR_VERSION,
+    }
+
+
+def _copilot_openai_pair(token: str) -> tuple[AsyncOpenAI, OpenAI]:
+    kwargs: dict[str, Any] = {
+        "api_key": token,
+        "base_url": _COPILOT_BASE_URL,
+        "max_retries": 0,
+        "default_headers": _copilot_default_headers(),
+    }
+    return AsyncOpenAI(**kwargs), OpenAI(**kwargs)
+
+
+def _require_copilot_env(_: Path | None = None) -> None:
+    _require_string(
+        _get_github_token(),
+        "No GitHub token found (set GH_TOKEN, GITHUB_TOKEN, or run `gh auth login`)",
+    )
+
+
+def _fetch_copilot_models_raw(token: str) -> list[JSONDict]:
+    """Fetch the full model metadata list from the Copilot API."""
+    resp = httpx.get(
+        f"{_COPILOT_BASE_URL}/models",
+        headers={
+            "Authorization": f"Bearer {token}",
+            **_copilot_default_headers(),
+        },
+        timeout=15,
+    )
+    resp.raise_for_status()
+    data = resp.json()
+    return data.get("data", []) if isinstance(data, dict) else []
+
+
+def _copilot_chat_model_ids(token: str) -> list[str]:
+    """Return sorted model IDs that support chat (not embeddings)."""
+    raw = _fetch_copilot_models_raw(token)
+    return sorted(
+        m["id"]
+        for m in raw
+        if isinstance(m.get("id"), str)
+        and m.get("capabilities", {}).get("type") == "chat"
+    )
+
+
+def _copilot_responses_model_ids(token: str) -> set[str]:
+    """Return model IDs that support the /responses endpoint."""
+    raw = _fetch_copilot_models_raw(token)
+    return {
+        m["id"]
+        for m in raw
+        if isinstance(m.get("id"), str)
+        and "/responses" in (m.get("supported_endpoints") or [])
+    }
+
+
+def _build_copilot_client(
+    region: str | None = None, cwd: Path | None = None
+) -> CompletionClient:
+    """Build a Copilot client that routes to /responses or /chat/completions."""
+    _ = region, cwd
+    token = _require_string(_get_github_token(), "No GitHub token found")
+    async_client, sync_client = _copilot_openai_pair(token)
+
+    # Probe which models support /responses vs /chat/completions
+    try:
+        responses_models = _copilot_responses_model_ids(token)
+    except Exception:
+        responses_models = set()
+
+    responses_inner = _openai_responses_client(
+        async_client, sync_client, fallback_models=None, default_models=None
+    )
+    chat_inner = _openai_chat_completions_client(
+        async_client, sync_client, tools_map=_tool_specs_to_openai
+    )
+
+    async def chat_completion(
+        model: str,
+        messages: list[ChatMessage],
+        tools: list[ToolSpec] | None = None,
+        tool_choice: str = "auto",
+        on_retry=None,
+    ) -> AssistantMessage:
+        inner = responses_inner if model in responses_models else chat_inner
+        return await inner.chat_completion(model, messages, tools, tool_choice, on_retry)
+
+    def list_models() -> list[str]:
+        try:
+            return _copilot_chat_model_ids(token)
+        except Exception:
+            return sorted(
+                m.id
+                for m in sync_client.models.list()
+                if not m.id.startswith("text-embedding")
+            )
+
+    return CompletionClient(chat_completion=chat_completion, list_models=list_models)
+
+
+def _list_copilot_models(
+    region: str | None = None, cwd: Path | None = None
+) -> list[str]:
+    _ = region, cwd
+    return _build_copilot_client(region, cwd).list_models()
+
+
 SHIM_SPECS: dict[str, ShimSpec] = {
     SHIM_OPENAI: ShimSpec(
         name=SHIM_OPENAI,
@@ -2521,6 +2679,12 @@ SHIM_SPECS: dict[str, ShimSpec] = {
         build_client=_build_claude_shim,
         list_models=_list_claude_models,
     ),
+    SHIM_COPILOT: ShimSpec(
+        name=SHIM_COPILOT,
+        ensure_env=_require_copilot_env,
+        build_client=_build_copilot_client,
+        list_models=_list_copilot_models,
+    ),
 }
 
 

{oy_cli-0.3.3 → oy_cli-0.3.5}/tests/test_oy_cli.py

@@ -220,5 +220,49 @@ class ClipTokensTests(unittest.TestCase):
         self.assertIn("tokens omitted", result)
 
 
+class JsonPathTests(unittest.TestCase):
+    """Tests for _json_path including the depth cap (M3)."""
+
+    def test_simple_dict_traversal(self):
+        self.assertEqual(oy_cli._json_path({"a": {"b": 1}}, "a.b"), 1)
+
+    def test_list_index(self):
+        self.assertEqual(oy_cli._json_path([10, 20, 30], "1"), 20)
+
+    def test_missing_key_raises(self):
+        with self.assertRaisesRegex(ValueError, "key not found"):
+            oy_cli._json_path({"a": 1}, "b")
+
+    def test_non_integer_index_raises(self):
+        with self.assertRaisesRegex(ValueError, "expected index"):
+            oy_cli._json_path([1, 2], "foo")
+
+    def test_depth_cap_raises(self):
+        # Build a deeply nested structure and a path exceeding 20 levels
+        obj = 0
+        for _ in range(25):
+            obj = {"a": obj}
+        deep = ".".join(["a"] * 21)
+        with self.assertRaisesRegex(ValueError, "json_path exceeded max depth"):
+            oy_cli._json_path(obj, deep)
+
+    def test_index_error_caught(self):
+        with self.assertRaisesRegex(ValueError, "out of range"):
+            oy_cli._json_path([1, 2], "5")
+
+    def test_empty_path_returns_value(self):
+        self.assertEqual(oy_cli._json_path(42, ""), 42)
+
+
+class CommandEnvImmutabilityTests(unittest.TestCase):
+    """Test that command_env returns an immutable mapping (L3)."""
+
+    def test_command_env_is_immutable(self):
+        from shim import command_env
+        env = command_env()
+        with self.assertRaises(TypeError):
+            env["SHOULD_NOT_WORK"] = "value"
+
+
 if __name__ == "__main__":
     unittest.main()