oy-cli 0.3.4__tar.gz → 0.4.0__tar.gz

{oy_cli-0.3.4 → oy_cli-0.4.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: oy-cli
-Version: 0.3.4
+Version: 0.4.0
 Summary: Tiny local coding CLI with a small tool surface
 Author: oy-cli contributors
 License-Expression: Apache-2.0
@@ -13,32 +13,30 @@ License-File: LICENSE
 Requires-Dist: boto3>=1.35.0
 Requires-Dist: defopt>=7.0.0
 Requires-Dist: httpx>=0.28.1
-Requires-Dist: markdownify>=1.2.0
 Requires-Dist: openai>=1.68.0
 Requires-Dist: rich>=14.3.0
 Requires-Dist: tiktoken>=0.7.0
 Requires-Dist: tenacity>=9.0.0
 Requires-Dist: msgspec>=0.20.0
+Requires-Dist: headroom-ai>=0.4.6
 Dynamic: license-file
 
 # oy-cli
 
 [![PyPI](https://img.shields.io/pypi/v/oy-cli)](https://pypi.org/project/oy-cli/)
 
-**Tiny AI coding assistant for your shell.** Reads files, runs commands, makes precise edits, and stays intentionally small.
+**AI coding assistant for your shell.** Reads files, searches content, and runs commands.
 
 ```bash
 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
 # Basic usage
-oy "read the main module and suggest improvements"
+oy "inspect the main module and suggest improvements"
 
 # Work in a specific directory
 OY_ROOT=./my-project oy "fix the failing tests"
@@ -63,29 +61,29 @@ oy --help                 # Show all commands
 
 ## Why This Exists
 
-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.
+`oy` is small, auditable, 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.
+new session each run, and explicit checkpoints when needed.
 
 ## System Prompts
 
-The system prompt is intentionally short. Tool semantics live with the tool definitions; the system prompt focuses on operating rules and judgment:
+The system prompt is short. Tool semantics live with the tool definitions; the system prompt focuses on operating rules and judgment:
 
 ### Base Prompt
 
 ```markdown
-You are oy, a tiny coding cli with tools.
-Work by inspecting first, then making explicit changes. Prefer simple auditable solutions.
-Keep going until done or genuinely blocked; if blocked, say what you tried and next steps.
-Use grugbrain-style simplicity for complexity, OWASP-minded judgment for security, and performance-aware judgment to avoid obvious waste.
+You are oy, a coding cli with tools.
+Inspect before editing with `read` for file content, `search` for regex matches, and `list` for path discovery. For existing code changes, prefer syntax-aware edits via `ast-grep` run through `bash`. Keep edits small, auditable, and verified with `read`, `git diff`, and batch independent tool calls.
+Keep going until done or blocked; if blocked, say what you tried and next steps.
+Use grugbrain.dev approach for maintainability/simplicity, OWASP-minded judgment for security, and performance-aware programming (Computer, Enhance!).
 ```
 
 ### Interactive Appendix
 
 ```markdown
-Use ask only when significant clarification or direction is needed.
+Use ask only when clarification or direction is needed.
 ```
 
 ### Non-Interactive Appendix
@@ -96,20 +94,21 @@ Non-interactive mode: do not pause for approval.
 
 ## Tools
 
-Each tool description is passed directly to the model. These are the exact descriptions:
+Each tool description is passed directly to the model:
 
 | Tool | Description |
 |------|-------------|
-| `list` | List a directory. Use this first on unfamiliar trees. Returns sorted entries, one per line, with / for directories. |
-| `read` | Read a file or directory. Use before editing. Files return line-numbered text; directories fall back to list. Use offset/limit for large files. |
-| `apply` | Edit files inside the workspace. Operations: replace, write, move, delete. Read first and keep edits precise. |
-| `bash` | Run shell commands for tests, builds, git, and scripts. Do not use for routine file inspection. Returns stdout and stderr together. |
-| `grep` | Search file contents by text or regex. Use file_glob to narrow by filename pattern. Returns matching lines with file and line numbers. |
-| `glob` | Find files by name pattern like '*.py' or 'src/**/*.js'. Use when you know the path shape. Supports *, ?, and **. |
-| `httpx` | Fetch web pages or APIs over HTTP(S). Presets: page, json, post_json. Use json_path to extract nested fields. Sensitive headers are redacted. |
-| `ask` | Ask the user a question in interactive runs. Use for significant ambiguity or decisions. Provide choices when useful. |
+| `list` | List paths by calling `Path.glob(path)`. Defaults to `path: "*"`. Use `src/*` or `src/**/*.py` exactly like pathlib glob patterns. Returns sorted entries, one per line, with / for directories. |
+| `read` | Read a file or directory. Files return line-numbered text. Directories return sorted entries, one per line, with / for directories. Use `offset` and `limit` for large files. |
+| `bash` | Shell commands are easy to run. For edits, prefer `ast-grep` for precise search/replace, `scc` for code-count analysis, and `xh` for web/API interaction; pipe to `rg` or `yq` for filtering when useful. These tools are effective for their niches, guaranteed to be available during an `oy` run, and their current usage docs can be checked with `--help`. For inspection, prefer the `search` tool. Returns structured results with `command`, `exit_code`, `ok`, `output_format`, `output`, and `truncated`. JSON output is parsed when possible. |
+| `search` | Search with ripgrep JSON output. Takes `pattern` and `path`, then passes any extra ripgrep flags from `args`, for example `pattern: 'needle', path: 'src', args: ['--glob', '*.py', '-i']`. `limit` only limits displayed results after ripgrep runs. |
+| `ask` | Ask the user a question in interactive runs. Use for ambiguity or decisions. Provide choices. |
+
+**Output truncation:** tool output is clipped to preserve context window; `bash` summarizes output into a single `output` field and marks truncation with one `truncated` flag. When clipped, narrow the next query or use `search` with a tighter `path` instead of re-running broad inspection.
+
+**Conversation compaction:** interactive chat compresses prepared context with [Headroom](https://github.com/chopratejas/headroom) before each model request, then falls back to omitting the oldest messages if the transcript still does not fit.
 
-**Output truncation:** tool output is clipped to preserve context window; `bash` keeps both head and tail. When clipped, narrow the next query.
+**Parallel tool calls:** `oy` can execute multiple tool calls returned in a single assistant turn. Explicit provider flags for parallel tool calls are only sent where the upstream API supports them directly today; other providers rely on their native tool-calling behavior.
 
 ## Audit Command
 
@@ -119,13 +118,14 @@ Each tool description is passed directly to the model. These are the exact descr
 
 ```markdown
 Audit the repo for security, unnecessary complexity, and major
-obvious performance issues, preserving project and human context.
+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
+the latest Git commit reference, and a codebase summary
+using tools like `scc`. 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
+using `bash` with `xh` (pipe to `rg` or `yq` if useful),
+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
@@ -146,7 +146,7 @@ OY_ROOT=./src oy audit      # Audit specific directory
 | Variable | Purpose |
 |----------|---------|
 | `OY_MODEL` | Override model for this session (bare name or `shim:model`) |
-| `OY_SHIM` | Force a specific shim: `openai`, `codex`, `gemini`, `claude`, `bedrock`, or `bedrock-mantle` |
+| `OY_SHIM` | Force a specific shim: `openai`, `codex`, `gemini`, `claude`, `copilot`, `bedrock`, or `bedrock-mantle` |
 | `OY_NON_INTERACTIVE` | Set to `1` to disable checkpoints |
 | `OY_ROOT` | Run against different workspace |
 | `OY_SYSTEM_FILE` | Append extra system instructions |
@@ -165,16 +165,17 @@ 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.
+**Model notes:** From testing, `glm-5` balances intelligence,
+cost, and tool-use ability. `kimi-k2.5` is another option.
 The [Artificial Analysis Comparison of Open Source Models](https://artificialanalysis.ai/models/open-source)
-is a good reference.
+is a reference.
 
 ## Requirements
 
 - Python 3.14+
 - `bash`
-- (Optional) `rg` (ripgrep) for faster search
+- `mise` installed and activated in the shell before launching `oy`
+- (Optional helper CLIs; `oy` auto-installs them on demand via `mise`): `rg` (ripgrep), `ast-grep`, `scc`, `xh`, `yq`
 - 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
@@ -199,10 +200,10 @@ export OPENAI_BASE_URL=https://your-endpoint.example/v1
 export OPENAI_API_KEY=...
 ```
 
-Gemini, Claude, Codex (OpenAI) creds should be automatically introspected
+Gemini, Claude, Copilot, and Codex (OpenAI) creds are 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.
+**AWS Bedrock:** Uses your default AWS profile/region. Supports auto-refresh of stale SSO sessions.
 ```bash
 export AWS_PROFILE=my-profile
 export AWS_REGION=us-west-2
@@ -218,6 +219,10 @@ ensure your profile has `bedrock:InvokeModel` permission.
 
 **"AWS SSO session is stale"** -> Run `aws sso login --use-device-code --no-browser`.
 
+**"Missing helper tool"** -> Install or activate `mise`, then rerun `oy`; `oy` assumes a working `mise` shell activation and auto-installs missing helper CLIs together through `mise`.
+
+**"`mise` is required; install and activate `mise` before running `oy`."** -> Install `mise`, activate it in your shell, then relaunch `oy`.
+
 ## Security
 
 `oy` can run shell commands and modify files with your permissions. Treat it like any other local automation tool.
@@ -228,9 +233,7 @@ 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.
+**Protections:** workspace-bound file access for built-in file tools and native boto3 credential resolution for Bedrock.
 
 ## Links
 

{oy_cli-0.3.4 → oy_cli-0.4.0}/README.md

@@ -2,20 +2,18 @@
 
 [![PyPI](https://img.shields.io/pypi/v/oy-cli)](https://pypi.org/project/oy-cli/)
 
-**Tiny AI coding assistant for your shell.** Reads files, runs commands, makes precise edits, and stays intentionally small.
+**AI coding assistant for your shell.** Reads files, searches content, and runs commands.
 
 ```bash
 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
 # Basic usage
-oy "read the main module and suggest improvements"
+oy "inspect the main module and suggest improvements"
 
 # Work in a specific directory
 OY_ROOT=./my-project oy "fix the failing tests"
@@ -40,29 +38,29 @@ oy --help                 # Show all commands
 
 ## Why This Exists
 
-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.
+`oy` is small, auditable, 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.
+new session each run, and explicit checkpoints when needed.
 
 ## System Prompts
 
-The system prompt is intentionally short. Tool semantics live with the tool definitions; the system prompt focuses on operating rules and judgment:
+The system prompt is short. Tool semantics live with the tool definitions; the system prompt focuses on operating rules and judgment:
 
 ### Base Prompt
 
 ```markdown
-You are oy, a tiny coding cli with tools.
-Work by inspecting first, then making explicit changes. Prefer simple auditable solutions.
-Keep going until done or genuinely blocked; if blocked, say what you tried and next steps.
-Use grugbrain-style simplicity for complexity, OWASP-minded judgment for security, and performance-aware judgment to avoid obvious waste.
+You are oy, a coding cli with tools.
+Inspect before editing with `read` for file content, `search` for regex matches, and `list` for path discovery. For existing code changes, prefer syntax-aware edits via `ast-grep` run through `bash`. Keep edits small, auditable, and verified with `read`, `git diff`, and batch independent tool calls.
+Keep going until done or blocked; if blocked, say what you tried and next steps.
+Use grugbrain.dev approach for maintainability/simplicity, OWASP-minded judgment for security, and performance-aware programming (Computer, Enhance!).
 ```
 
 ### Interactive Appendix
 
 ```markdown
-Use ask only when significant clarification or direction is needed.
+Use ask only when clarification or direction is needed.
 ```
 
 ### Non-Interactive Appendix
@@ -73,20 +71,21 @@ Non-interactive mode: do not pause for approval.
 
 ## Tools
 
-Each tool description is passed directly to the model. These are the exact descriptions:
+Each tool description is passed directly to the model:
 
 | Tool | Description |
 |------|-------------|
-| `list` | List a directory. Use this first on unfamiliar trees. Returns sorted entries, one per line, with / for directories. |
-| `read` | Read a file or directory. Use before editing. Files return line-numbered text; directories fall back to list. Use offset/limit for large files. |
-| `apply` | Edit files inside the workspace. Operations: replace, write, move, delete. Read first and keep edits precise. |
-| `bash` | Run shell commands for tests, builds, git, and scripts. Do not use for routine file inspection. Returns stdout and stderr together. |
-| `grep` | Search file contents by text or regex. Use file_glob to narrow by filename pattern. Returns matching lines with file and line numbers. |
-| `glob` | Find files by name pattern like '*.py' or 'src/**/*.js'. Use when you know the path shape. Supports *, ?, and **. |
-| `httpx` | Fetch web pages or APIs over HTTP(S). Presets: page, json, post_json. Use json_path to extract nested fields. Sensitive headers are redacted. |
-| `ask` | Ask the user a question in interactive runs. Use for significant ambiguity or decisions. Provide choices when useful. |
+| `list` | List paths by calling `Path.glob(path)`. Defaults to `path: "*"`. Use `src/*` or `src/**/*.py` exactly like pathlib glob patterns. Returns sorted entries, one per line, with / for directories. |
+| `read` | Read a file or directory. Files return line-numbered text. Directories return sorted entries, one per line, with / for directories. Use `offset` and `limit` for large files. |
+| `bash` | Shell commands are easy to run. For edits, prefer `ast-grep` for precise search/replace, `scc` for code-count analysis, and `xh` for web/API interaction; pipe to `rg` or `yq` for filtering when useful. These tools are effective for their niches, guaranteed to be available during an `oy` run, and their current usage docs can be checked with `--help`. For inspection, prefer the `search` tool. Returns structured results with `command`, `exit_code`, `ok`, `output_format`, `output`, and `truncated`. JSON output is parsed when possible. |
+| `search` | Search with ripgrep JSON output. Takes `pattern` and `path`, then passes any extra ripgrep flags from `args`, for example `pattern: 'needle', path: 'src', args: ['--glob', '*.py', '-i']`. `limit` only limits displayed results after ripgrep runs. |
+| `ask` | Ask the user a question in interactive runs. Use for ambiguity or decisions. Provide choices. |
+
+**Output truncation:** tool output is clipped to preserve context window; `bash` summarizes output into a single `output` field and marks truncation with one `truncated` flag. When clipped, narrow the next query or use `search` with a tighter `path` instead of re-running broad inspection.
+
+**Conversation compaction:** interactive chat compresses prepared context with [Headroom](https://github.com/chopratejas/headroom) before each model request, then falls back to omitting the oldest messages if the transcript still does not fit.
 
-**Output truncation:** tool output is clipped to preserve context window; `bash` keeps both head and tail. When clipped, narrow the next query.
+**Parallel tool calls:** `oy` can execute multiple tool calls returned in a single assistant turn. Explicit provider flags for parallel tool calls are only sent where the upstream API supports them directly today; other providers rely on their native tool-calling behavior.
 
 ## Audit Command
 
@@ -96,13 +95,14 @@ Each tool description is passed directly to the model. These are the exact descr
 
 ```markdown
 Audit the repo for security, unnecessary complexity, and major
-obvious performance issues, preserving project and human context.
+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
+the latest Git commit reference, and a codebase summary
+using tools like `scc`. 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
+using `bash` with `xh` (pipe to `rg` or `yq` if useful),
+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
@@ -123,7 +123,7 @@ OY_ROOT=./src oy audit      # Audit specific directory
 | Variable | Purpose |
 |----------|---------|
 | `OY_MODEL` | Override model for this session (bare name or `shim:model`) |
-| `OY_SHIM` | Force a specific shim: `openai`, `codex`, `gemini`, `claude`, `bedrock`, or `bedrock-mantle` |
+| `OY_SHIM` | Force a specific shim: `openai`, `codex`, `gemini`, `claude`, `copilot`, `bedrock`, or `bedrock-mantle` |
 | `OY_NON_INTERACTIVE` | Set to `1` to disable checkpoints |
 | `OY_ROOT` | Run against different workspace |
 | `OY_SYSTEM_FILE` | Append extra system instructions |
@@ -142,16 +142,17 @@ 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.
+**Model notes:** From testing, `glm-5` balances intelligence,
+cost, and tool-use ability. `kimi-k2.5` is another option.
 The [Artificial Analysis Comparison of Open Source Models](https://artificialanalysis.ai/models/open-source)
-is a good reference.
+is a reference.
 
 ## Requirements
 
 - Python 3.14+
 - `bash`
-- (Optional) `rg` (ripgrep) for faster search
+- `mise` installed and activated in the shell before launching `oy`
+- (Optional helper CLIs; `oy` auto-installs them on demand via `mise`): `rg` (ripgrep), `ast-grep`, `scc`, `xh`, `yq`
 - 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
@@ -176,10 +177,10 @@ export OPENAI_BASE_URL=https://your-endpoint.example/v1
 export OPENAI_API_KEY=...
 ```
 
-Gemini, Claude, Codex (OpenAI) creds should be automatically introspected
+Gemini, Claude, Copilot, and Codex (OpenAI) creds are 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.
+**AWS Bedrock:** Uses your default AWS profile/region. Supports auto-refresh of stale SSO sessions.
 ```bash
 export AWS_PROFILE=my-profile
 export AWS_REGION=us-west-2
@@ -195,6 +196,10 @@ ensure your profile has `bedrock:InvokeModel` permission.
 
 **"AWS SSO session is stale"** -> Run `aws sso login --use-device-code --no-browser`.
 
+**"Missing helper tool"** -> Install or activate `mise`, then rerun `oy`; `oy` assumes a working `mise` shell activation and auto-installs missing helper CLIs together through `mise`.
+
+**"`mise` is required; install and activate `mise` before running `oy`."** -> Install `mise`, activate it in your shell, then relaunch `oy`.
+
 ## Security
 
 `oy` can run shell commands and modify files with your permissions. Treat it like any other local automation tool.
@@ -205,9 +210,7 @@ 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.
+**Protections:** workspace-bound file access for built-in file tools and native boto3 credential resolution for Bedrock.
 
 ## Links
 

{oy_cli-0.3.4 → oy_cli-0.4.0}/oy_cli.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: oy-cli
-Version: 0.3.4
+Version: 0.4.0
 Summary: Tiny local coding CLI with a small tool surface
 Author: oy-cli contributors
 License-Expression: Apache-2.0
@@ -13,32 +13,30 @@ License-File: LICENSE
 Requires-Dist: boto3>=1.35.0
 Requires-Dist: defopt>=7.0.0
 Requires-Dist: httpx>=0.28.1
-Requires-Dist: markdownify>=1.2.0
 Requires-Dist: openai>=1.68.0
 Requires-Dist: rich>=14.3.0
 Requires-Dist: tiktoken>=0.7.0
 Requires-Dist: tenacity>=9.0.0
 Requires-Dist: msgspec>=0.20.0
+Requires-Dist: headroom-ai>=0.4.6
 Dynamic: license-file
 
 # oy-cli
 
 [![PyPI](https://img.shields.io/pypi/v/oy-cli)](https://pypi.org/project/oy-cli/)
 
-**Tiny AI coding assistant for your shell.** Reads files, runs commands, makes precise edits, and stays intentionally small.
+**AI coding assistant for your shell.** Reads files, searches content, and runs commands.
 
 ```bash
 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
 # Basic usage
-oy "read the main module and suggest improvements"
+oy "inspect the main module and suggest improvements"
 
 # Work in a specific directory
 OY_ROOT=./my-project oy "fix the failing tests"
@@ -63,29 +61,29 @@ oy --help                 # Show all commands
 
 ## Why This Exists
 
-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.
+`oy` is small, auditable, 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.
+new session each run, and explicit checkpoints when needed.
 
 ## System Prompts
 
-The system prompt is intentionally short. Tool semantics live with the tool definitions; the system prompt focuses on operating rules and judgment:
+The system prompt is short. Tool semantics live with the tool definitions; the system prompt focuses on operating rules and judgment:
 
 ### Base Prompt
 
 ```markdown
-You are oy, a tiny coding cli with tools.
-Work by inspecting first, then making explicit changes. Prefer simple auditable solutions.
-Keep going until done or genuinely blocked; if blocked, say what you tried and next steps.
-Use grugbrain-style simplicity for complexity, OWASP-minded judgment for security, and performance-aware judgment to avoid obvious waste.
+You are oy, a coding cli with tools.
+Inspect before editing with `read` for file content, `search` for regex matches, and `list` for path discovery. For existing code changes, prefer syntax-aware edits via `ast-grep` run through `bash`. Keep edits small, auditable, and verified with `read`, `git diff`, and batch independent tool calls.
+Keep going until done or blocked; if blocked, say what you tried and next steps.
+Use grugbrain.dev approach for maintainability/simplicity, OWASP-minded judgment for security, and performance-aware programming (Computer, Enhance!).
 ```
 
 ### Interactive Appendix
 
 ```markdown
-Use ask only when significant clarification or direction is needed.
+Use ask only when clarification or direction is needed.
 ```
 
 ### Non-Interactive Appendix
@@ -96,20 +94,21 @@ Non-interactive mode: do not pause for approval.
 
 ## Tools
 
-Each tool description is passed directly to the model. These are the exact descriptions:
+Each tool description is passed directly to the model:
 
 | Tool | Description |
 |------|-------------|
-| `list` | List a directory. Use this first on unfamiliar trees. Returns sorted entries, one per line, with / for directories. |
-| `read` | Read a file or directory. Use before editing. Files return line-numbered text; directories fall back to list. Use offset/limit for large files. |
-| `apply` | Edit files inside the workspace. Operations: replace, write, move, delete. Read first and keep edits precise. |
-| `bash` | Run shell commands for tests, builds, git, and scripts. Do not use for routine file inspection. Returns stdout and stderr together. |
-| `grep` | Search file contents by text or regex. Use file_glob to narrow by filename pattern. Returns matching lines with file and line numbers. |
-| `glob` | Find files by name pattern like '*.py' or 'src/**/*.js'. Use when you know the path shape. Supports *, ?, and **. |
-| `httpx` | Fetch web pages or APIs over HTTP(S). Presets: page, json, post_json. Use json_path to extract nested fields. Sensitive headers are redacted. |
-| `ask` | Ask the user a question in interactive runs. Use for significant ambiguity or decisions. Provide choices when useful. |
+| `list` | List paths by calling `Path.glob(path)`. Defaults to `path: "*"`. Use `src/*` or `src/**/*.py` exactly like pathlib glob patterns. Returns sorted entries, one per line, with / for directories. |
+| `read` | Read a file or directory. Files return line-numbered text. Directories return sorted entries, one per line, with / for directories. Use `offset` and `limit` for large files. |
+| `bash` | Shell commands are easy to run. For edits, prefer `ast-grep` for precise search/replace, `scc` for code-count analysis, and `xh` for web/API interaction; pipe to `rg` or `yq` for filtering when useful. These tools are effective for their niches, guaranteed to be available during an `oy` run, and their current usage docs can be checked with `--help`. For inspection, prefer the `search` tool. Returns structured results with `command`, `exit_code`, `ok`, `output_format`, `output`, and `truncated`. JSON output is parsed when possible. |
+| `search` | Search with ripgrep JSON output. Takes `pattern` and `path`, then passes any extra ripgrep flags from `args`, for example `pattern: 'needle', path: 'src', args: ['--glob', '*.py', '-i']`. `limit` only limits displayed results after ripgrep runs. |
+| `ask` | Ask the user a question in interactive runs. Use for ambiguity or decisions. Provide choices. |
+
+**Output truncation:** tool output is clipped to preserve context window; `bash` summarizes output into a single `output` field and marks truncation with one `truncated` flag. When clipped, narrow the next query or use `search` with a tighter `path` instead of re-running broad inspection.
+
+**Conversation compaction:** interactive chat compresses prepared context with [Headroom](https://github.com/chopratejas/headroom) before each model request, then falls back to omitting the oldest messages if the transcript still does not fit.
 
-**Output truncation:** tool output is clipped to preserve context window; `bash` keeps both head and tail. When clipped, narrow the next query.
+**Parallel tool calls:** `oy` can execute multiple tool calls returned in a single assistant turn. Explicit provider flags for parallel tool calls are only sent where the upstream API supports them directly today; other providers rely on their native tool-calling behavior.
 
 ## Audit Command
 
@@ -119,13 +118,14 @@ Each tool description is passed directly to the model. These are the exact descr
 
 ```markdown
 Audit the repo for security, unnecessary complexity, and major
-obvious performance issues, preserving project and human context.
+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
+the latest Git commit reference, and a codebase summary
+using tools like `scc`. 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
+using `bash` with `xh` (pipe to `rg` or `yq` if useful),
+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
@@ -146,7 +146,7 @@ OY_ROOT=./src oy audit      # Audit specific directory
 | Variable | Purpose |
 |----------|---------|
 | `OY_MODEL` | Override model for this session (bare name or `shim:model`) |
-| `OY_SHIM` | Force a specific shim: `openai`, `codex`, `gemini`, `claude`, `bedrock`, or `bedrock-mantle` |
+| `OY_SHIM` | Force a specific shim: `openai`, `codex`, `gemini`, `claude`, `copilot`, `bedrock`, or `bedrock-mantle` |
 | `OY_NON_INTERACTIVE` | Set to `1` to disable checkpoints |
 | `OY_ROOT` | Run against different workspace |
 | `OY_SYSTEM_FILE` | Append extra system instructions |
@@ -165,16 +165,17 @@ 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.
+**Model notes:** From testing, `glm-5` balances intelligence,
+cost, and tool-use ability. `kimi-k2.5` is another option.
 The [Artificial Analysis Comparison of Open Source Models](https://artificialanalysis.ai/models/open-source)
-is a good reference.
+is a reference.
 
 ## Requirements
 
 - Python 3.14+
 - `bash`
-- (Optional) `rg` (ripgrep) for faster search
+- `mise` installed and activated in the shell before launching `oy`
+- (Optional helper CLIs; `oy` auto-installs them on demand via `mise`): `rg` (ripgrep), `ast-grep`, `scc`, `xh`, `yq`
 - 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
@@ -199,10 +200,10 @@ export OPENAI_BASE_URL=https://your-endpoint.example/v1
 export OPENAI_API_KEY=...
 ```
 
-Gemini, Claude, Codex (OpenAI) creds should be automatically introspected
+Gemini, Claude, Copilot, and Codex (OpenAI) creds are 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.
+**AWS Bedrock:** Uses your default AWS profile/region. Supports auto-refresh of stale SSO sessions.
 ```bash
 export AWS_PROFILE=my-profile
 export AWS_REGION=us-west-2
@@ -218,6 +219,10 @@ ensure your profile has `bedrock:InvokeModel` permission.
 
 **"AWS SSO session is stale"** -> Run `aws sso login --use-device-code --no-browser`.
 
+**"Missing helper tool"** -> Install or activate `mise`, then rerun `oy`; `oy` assumes a working `mise` shell activation and auto-installs missing helper CLIs together through `mise`.
+
+**"`mise` is required; install and activate `mise` before running `oy`."** -> Install `mise`, activate it in your shell, then relaunch `oy`.
+
 ## Security
 
 `oy` can run shell commands and modify files with your permissions. Treat it like any other local automation tool.
@@ -228,9 +233,7 @@ 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.
+**Protections:** workspace-bound file access for built-in file tools and native boto3 credential resolution for Bedrock.
 
 ## Links
 

{oy_cli-0.3.4 → oy_cli-0.4.0}/oy_cli.egg-info/SOURCES.txt

@@ -1,6 +1,7 @@
 LICENSE
 README.md
 oy_cli.py
+providers.py
 pyproject.toml
 shim.py
 oy_cli.egg-info/PKG-INFO
@@ -9,5 +10,6 @@ oy_cli.egg-info/dependency_links.txt
 oy_cli.egg-info/entry_points.txt
 oy_cli.egg-info/requires.txt
 oy_cli.egg-info/top_level.txt
+tests/test_async_cleanup.py
 tests/test_oy_cli.py
 tests/test_shim.py

{oy_cli-0.3.4 → oy_cli-0.4.0}/oy_cli.egg-info/requires.txt

@@ -1,9 +1,9 @@
 boto3>=1.35.0
 defopt>=7.0.0
 httpx>=0.28.1
-markdownify>=1.2.0
 openai>=1.68.0
 rich>=14.3.0
 tiktoken>=0.7.0
 tenacity>=9.0.0
 msgspec>=0.20.0
+headroom-ai>=0.4.6

{oy_cli-0.3.4 → oy_cli-0.4.0}/oy_cli.egg-info/top_level.txt

@@ -1,2 +1,3 @@
 oy_cli
+providers
 shim