git-cai-cli 0.11.2__tar.gz → 0.12.1__tar.gz

{git_cai_cli-0.11.2 → git_cai_cli-0.12.1}/.github/ci/cai_config.ci.yml

@@ -3,9 +3,12 @@ language: en
 style: professional
 emoji: false
 load_tokens_from: /home/runner/work/.config/cai/tokens.yml
+timeout: 30
+full_files: false
 anthropic:
   model: claude-haiku-4-5
   temperature: 0
+  max_tokens: 32768
 deepseek:
   model: deepseek-chat
   temperature: 0
@@ -18,6 +21,10 @@ groq:
 mistral:
   model: codestral-2508
   temperature: 0
+ollama:
+  model: llama3.1
+  temperature: 0
+  timeout: 300
 openai:
   model: gpt-5.1
   temperature: 0

{git_cai_cli-0.11.2 → git_cai_cli-0.12.1}/CLAUDE.md

@@ -9,8 +9,9 @@ This file provides guidance to Claude Code (claude.ai/code) when working with co
 ## Common Commands
 
 ```bash
+uv sync --dev          # Install dependencies (including dev deps)
 make test              # Run all tests (uv run pytest)
-make lint              # Branch name check + MegaLinter
+make lint              # Branch name check + MegaLinter (requires npx + Docker)
 make lint-fix          # Auto-fix lint issues
 make clean             # Clean uv cache and .venv
 
@@ -34,16 +35,19 @@ Entry point: `src/git_cai_cli/cli/cli.py` (Typer app) → `main.py` (dispatcher)
 
 ### Core Modules (core/)
 
-- **config.py** — Config loading with precedence: repo `cai_config.yml` > home `~/.config/cai/cai_config.yml` > bundled defaults
-- **llm.py** — `CommitMessageGenerator` class with `_dispatch_generate()` routing to provider-specific methods (`generate_openai()`, `generate_anthropic()`, etc.)
+- **config.py** — Config loading. Repo config is **authoritative** (not merged with home config): repo `cai_config.yml` > home `~/.config/cai/cai_config.yml` > bundled `DEFAULT_CONFIG`
+- **llm.py** — `CommitMessageGenerator` class with `_dispatch_generate()` routing to
+  provider-specific methods. OpenAI/DeepSeek use the OpenAI SDK;
+  Anthropic/Gemini/Groq/xAI/Mistral use direct HTTP via `requests`; Ollama uses its
+  local HTTP API and auto-starts the server if needed
 - **gitutils.py** — Git operations (find root, diff with `.caiignore` support, commit)
-- **validate.py** — Config and LLM call validation
+- **validate.py** — Config key/structure validation, language/style validation, and LLM call wrapper that converts auth errors to user-facing messages
 - **squash.py** — Squash commit workflow (stage → generate summary → editor → reset + commit)
 - **options.py** — `CliManager` orchestrating CLI operations
 
 ### Prompt Fallback Chain
 
-File (repo-level) → File (user home) → Bundled package (`defaults/*.md`) → Hardcoded string (`prompts_fallback.py`)
+File (repo-level) → File (user home `~/.config/cai/`) → Hardcoded string (`prompts_fallback.py`)
 
 ### Tokens
 
@@ -53,10 +57,19 @@ API tokens loaded from `~/.config/cai/tokens.yml`. Ollama is in `TOKENLESS_PROVI
 
 - Unit tests in `tests/unit/`, integration tests in `tests/integration/`
 - Uses `requests-mock` for HTTP mocking
-- CI tests against Python 3.10–3.14
+- `tests/conftest.py` inserts `src/` into `sys.path` so local source wins over any installed package
+- CI tests against Python 3.10–3.14; CI copies `.github/ci/cai_config.ci.yml` and `tokens.ci.yml` to `~/.config/cai/`
 
 ## Build & Versioning
 
 - Package manager: **uv**
 - Version: auto-generated from git tags via `setuptools-scm` → `src/git_cai_cli/_version.py`
 - Do not edit `_version.py` manually
+
+## Branch Naming
+
+Enforced by CI and `make lint` via `.linters/check_git_branch_name.sh`:
+
+```text
+master | (feature|fix|hotfix|chore|refactor|test|docs)/<lowercase-slug>
+```

{git_cai_cli-0.11.2/src/git_cai_cli.egg-info → git_cai_cli-0.12.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: git-cai-cli
-Version: 0.11.2
+Version: 0.12.1
 Summary: Use LLM to create git commit messages
 Author-email: Thorsten Foltz <thorsten.foltz@live.com>
 License-Expression: MIT
@@ -75,19 +75,18 @@ Currently supported providers:
 - Generates meaningful, context-aware commit messages using an LLM
 - Seamless integration with Git
 - Supports multiple LLM providers and models
-- Override provider and model per invocation (`-P`, `-m`)
 - Global configuration with per-repository overrides
 - Repository-specific language, style, and model selection
-- Amend the last commit message with a regenerated one (`-A`)
-- Conventional Commits format support (`-C`)
-- Change configuration from the command line (`-S`, `-H`)
+- Amend the last commit message with a regenerated one
+- Conventional Commits format support
+- Change configuration from the command line
 - Optional commit squashing with automatic summary generation (all, last N, or up to a specific commit)
-- List providers, models, active config, and file paths (`-l`)
+- List providers, models, active config, and file paths
 - Token usage logging for API calls
-- Branch name as LLM context via `--branch` / `-b`
-- Extra context for the LLM via `--context` / `-x` (e.g. ticket numbers, reasons)
-- Generation time measurement (`-t`)
-- Shell completion for bash, zsh, and fish (`-i`)
+- Branch name as LLM context
+- Extra context for the LLM
+- Generation time measurement
+- Shell completion for bash, zsh, and fish
 
 ---
 
@@ -124,7 +123,7 @@ Once installed, cai works like a standard Git command:
 git cai
 ```
 
-cai uses the output of `git diff` to generate a commit message.  
+cai uses the output of `git diff` to generate a commit message or, if optional, with the complete file content.
 The generated message is opened in your configured Git editor, allowing you to review or edit it before committing.
 
 In short: it behaves like `git commit`, but the commit message is pre-filled.
@@ -150,18 +149,21 @@ On first execution, cai automatically creates the base configuration in your hom
 - API tokens:  
   ~/.config/cai/tokens.yml
 
-It also creates two Markdown prompt files:
+It also creates three Markdown prompt files:
 
 - Default commit prompt:  
   ~/.config/cai/commit_prompt.md
 - Default squash prompt:  
   ~/.config/cai/squash_prompt.md
+- Default full-files prompt (used with `-F` / `--full-files`):  
+  ~/.config/cai/full_files_prompt.md
 
 Don't be scared the first run will show an error. It only misses a token.
 Add your provider API keys to `tokens.yml`. Once configured, cai will reuse them automatically.
+Optional for each repository a file containing tokens can be set.
 Set your preferred LLM in `cai_config.yml` (Groq by default).
 
-If you want to use Ollama, set `default: ollama` and configure the `ollama:` block (model/temperature). Ollama is automatically started when used.
+If you want to use Ollama, install it, set `default: ollama` and configure the `ollama:` block (model/temperature). Ollama is automatically started when used.
 
 ### Custom prompts (Markdown)
 
@@ -178,8 +180,12 @@ This creates:
 
 - `commit_prompt.md`
 - `squash_prompt.md`
+- `full_files_prompt.md`
+  so the LLM knows the full working-tree contents of each changed file are
+  attached and can reason about *why* each edit was made, not just what
+  the diff shows.
 
-Then set `prompt_file` and/or `squash_prompt_file` in your `cai_config.yml` (also repo) to point to those files.
+Then set `prompt_file`, `squash_prompt_file`, and/or `full_files_prompt_file` in your `cai_config.yml` (also repo) to point to those files.
 
 ### Repository-specific configuration
 
@@ -220,10 +226,13 @@ git cai -g
 - `load_tokens_from` – path to the file where API tokens are stored
 - `prompt_file` - path to the file where the prompt for the commit is stored
 - `squash_prompt_file` - path to the file where the prompt for the squash is stored
-- `branch_context` – include current branch name as LLM context (default: `false`)
-- `conventional` – use Conventional Commits format (default: `false`)
-- `token_logging` – log token usage after each LLM call (default: `true` for new installs)
-- `measure_time` – log generation time (default: `false`)
+- `full_files_prompt_file` - path to the prompt used when `-F` / `--full-files` attaches full file contents
+- `full_files` – attach always the full working-tree contents of affected files alongside the diff
+- `timeout` – HTTP timeout for LLM calls in seconds
+- `branch_context` – include current branch name as LLM context
+- `conventional` – use Conventional Commits format
+- `token_logging` – log token usage after each LLM call
+- `measure_time` – log generation time
 
 ---
 
@@ -237,6 +246,8 @@ In addition to `git cai`, the following options are available:
 - `-C`, `--conventional` – use Conventional Commits format (`type(scope): description`)
 - `-c`, `--crazy` – Trust the LLM and commit without checking
 - `-d`, `--debug` – enable debug logging
+- `-F`, `--full-files` – attach the full contents of affected files alongside the diff (uses `full_files_prompt.md`)
+- `-f`, `--files` `PATH` – limit the diff (and full-file content, if enabled) to PATH; repeat for multiple files
 - `-g`, `--generate-config` – generate the default `cai_config.yml` in the current directory
 - `-H`, `--set-home` – set a config value in home config (`key=value`), always targets `~/.config/cai/`
 - `-h`, `--help` – show help and available commands
@@ -246,13 +257,15 @@ In addition to `git cai`, the following options are available:
 - `-p`, `--generate-prompts` – generate default `commit_prompt.md` and `squash_prompt.md` in the current directory (for customization)
 - `-P`, `--provider` – override the LLM provider for this invocation
 - `-S`, `--set` – set a config value (`key=value`) in repo config (requires existing repo config)
-- `-H`, `--set-home` – set a config value in home config (`key=value`), always targets `~/.config/cai/`
 - `-s`, `--squash` `[N|HASH]` – squash commits on the current branch and summarize them. Without argument: squash all since branch checkout. With a number: squash the last N commits. With a commit hash: squash up to and including that commit
+- `-T`, `--timeout` `SECONDS` – HTTP timeout for this invocation (overrides config)
 - `-t`, `--time` – measure and log commit message generation time
 - `-x`, `--context` – provide extra context for the LLM (e.g. ticket number, reason for change)
 - `-u`, `--update` – check for updates
 - `-v`, `--version` – show the installed version
 
+## Examples
+
 ### Amend
 
 To regenerate the last commit message and amend it:
@@ -280,6 +293,30 @@ To enable it permanently:
 git cai -S conventional=true
 ```
 
+### Attaching full file contents and restricting to specific files
+
+Sometimes the diff alone is too little context for the LLM to explain *why* a
+change was made. `-F` / `--full-files` attaches the complete working-tree
+contents of every staged file alongside the diff and switches to a dedicated
+prompt (`full_files_prompt.md`) that instructs the LLM to use that full
+context to infer intent rather than just describe the mechanical edit.
+
+```sh
+git cai -F                         # attach full contents of all staged files
+git cai -F -f src/foo.py -f src/bar.py   # only these files
+git cai -f docs/README.md          # restrict diff to one file (no full files)
+```
+
+Both flags log the affected files at INFO level as paths relative to the
+repository root, so you can see exactly what ends up in the prompt — binaries,
+deleted files, and paths matched by `.caiignore` are filtered out and skipped.
+
+Persist the default:
+
+```sh
+git cai -S full_files=true
+```
+
 ### Changing configuration from the CLI
 
 Instead of editing YAML files manually, use `--set` or `--set-home` to update config values.
@@ -288,7 +325,7 @@ Instead of editing YAML files manually, use `--set` or `--set-home` to update co
 
 ```sh
 git cai -S default=anthropic           # change the default provider
-git cai -S emoji=false                  # disable emojis
+git cai -S emoji=false                 # disable emojis
 git cai -S groq.model=llama-3.3-70b    # nested key (dot notation)
 git cai -S openai.temperature=0.7      # set temperature as float
 ```

{git_cai_cli-0.11.2 → git_cai_cli-0.12.1}/README.md

@@ -46,19 +46,18 @@ Currently supported providers:
 - Generates meaningful, context-aware commit messages using an LLM
 - Seamless integration with Git
 - Supports multiple LLM providers and models
-- Override provider and model per invocation (`-P`, `-m`)
 - Global configuration with per-repository overrides
 - Repository-specific language, style, and model selection
-- Amend the last commit message with a regenerated one (`-A`)
-- Conventional Commits format support (`-C`)
-- Change configuration from the command line (`-S`, `-H`)
+- Amend the last commit message with a regenerated one
+- Conventional Commits format support
+- Change configuration from the command line
 - Optional commit squashing with automatic summary generation (all, last N, or up to a specific commit)
-- List providers, models, active config, and file paths (`-l`)
+- List providers, models, active config, and file paths
 - Token usage logging for API calls
-- Branch name as LLM context via `--branch` / `-b`
-- Extra context for the LLM via `--context` / `-x` (e.g. ticket numbers, reasons)
-- Generation time measurement (`-t`)
-- Shell completion for bash, zsh, and fish (`-i`)
+- Branch name as LLM context
+- Extra context for the LLM
+- Generation time measurement
+- Shell completion for bash, zsh, and fish
 
 ---
 
@@ -95,7 +94,7 @@ Once installed, cai works like a standard Git command:
 git cai
 ```
 
-cai uses the output of `git diff` to generate a commit message.  
+cai uses the output of `git diff` to generate a commit message or, if optional, with the complete file content.
 The generated message is opened in your configured Git editor, allowing you to review or edit it before committing.
 
 In short: it behaves like `git commit`, but the commit message is pre-filled.
@@ -121,18 +120,21 @@ On first execution, cai automatically creates the base configuration in your hom
 - API tokens:  
   ~/.config/cai/tokens.yml
 
-It also creates two Markdown prompt files:
+It also creates three Markdown prompt files:
 
 - Default commit prompt:  
   ~/.config/cai/commit_prompt.md
 - Default squash prompt:  
   ~/.config/cai/squash_prompt.md
+- Default full-files prompt (used with `-F` / `--full-files`):  
+  ~/.config/cai/full_files_prompt.md
 
 Don't be scared the first run will show an error. It only misses a token.
 Add your provider API keys to `tokens.yml`. Once configured, cai will reuse them automatically.
+Optional for each repository a file containing tokens can be set.
 Set your preferred LLM in `cai_config.yml` (Groq by default).
 
-If you want to use Ollama, set `default: ollama` and configure the `ollama:` block (model/temperature). Ollama is automatically started when used.
+If you want to use Ollama, install it, set `default: ollama` and configure the `ollama:` block (model/temperature). Ollama is automatically started when used.
 
 ### Custom prompts (Markdown)
 
@@ -149,8 +151,12 @@ This creates:
 
 - `commit_prompt.md`
 - `squash_prompt.md`
+- `full_files_prompt.md`
+  so the LLM knows the full working-tree contents of each changed file are
+  attached and can reason about *why* each edit was made, not just what
+  the diff shows.
 
-Then set `prompt_file` and/or `squash_prompt_file` in your `cai_config.yml` (also repo) to point to those files.
+Then set `prompt_file`, `squash_prompt_file`, and/or `full_files_prompt_file` in your `cai_config.yml` (also repo) to point to those files.
 
 ### Repository-specific configuration
 
@@ -191,10 +197,13 @@ git cai -g
 - `load_tokens_from` – path to the file where API tokens are stored
 - `prompt_file` - path to the file where the prompt for the commit is stored
 - `squash_prompt_file` - path to the file where the prompt for the squash is stored
-- `branch_context` – include current branch name as LLM context (default: `false`)
-- `conventional` – use Conventional Commits format (default: `false`)
-- `token_logging` – log token usage after each LLM call (default: `true` for new installs)
-- `measure_time` – log generation time (default: `false`)
+- `full_files_prompt_file` - path to the prompt used when `-F` / `--full-files` attaches full file contents
+- `full_files` – attach always the full working-tree contents of affected files alongside the diff
+- `timeout` – HTTP timeout for LLM calls in seconds
+- `branch_context` – include current branch name as LLM context
+- `conventional` – use Conventional Commits format
+- `token_logging` – log token usage after each LLM call
+- `measure_time` – log generation time
 
 ---
 
@@ -208,6 +217,8 @@ In addition to `git cai`, the following options are available:
 - `-C`, `--conventional` – use Conventional Commits format (`type(scope): description`)
 - `-c`, `--crazy` – Trust the LLM and commit without checking
 - `-d`, `--debug` – enable debug logging
+- `-F`, `--full-files` – attach the full contents of affected files alongside the diff (uses `full_files_prompt.md`)
+- `-f`, `--files` `PATH` – limit the diff (and full-file content, if enabled) to PATH; repeat for multiple files
 - `-g`, `--generate-config` – generate the default `cai_config.yml` in the current directory
 - `-H`, `--set-home` – set a config value in home config (`key=value`), always targets `~/.config/cai/`
 - `-h`, `--help` – show help and available commands
@@ -217,13 +228,15 @@ In addition to `git cai`, the following options are available:
 - `-p`, `--generate-prompts` – generate default `commit_prompt.md` and `squash_prompt.md` in the current directory (for customization)
 - `-P`, `--provider` – override the LLM provider for this invocation
 - `-S`, `--set` – set a config value (`key=value`) in repo config (requires existing repo config)
-- `-H`, `--set-home` – set a config value in home config (`key=value`), always targets `~/.config/cai/`
 - `-s`, `--squash` `[N|HASH]` – squash commits on the current branch and summarize them. Without argument: squash all since branch checkout. With a number: squash the last N commits. With a commit hash: squash up to and including that commit
+- `-T`, `--timeout` `SECONDS` – HTTP timeout for this invocation (overrides config)
 - `-t`, `--time` – measure and log commit message generation time
 - `-x`, `--context` – provide extra context for the LLM (e.g. ticket number, reason for change)
 - `-u`, `--update` – check for updates
 - `-v`, `--version` – show the installed version
 
+## Examples
+
 ### Amend
 
 To regenerate the last commit message and amend it:
@@ -251,6 +264,30 @@ To enable it permanently:
 git cai -S conventional=true
 ```
 
+### Attaching full file contents and restricting to specific files
+
+Sometimes the diff alone is too little context for the LLM to explain *why* a
+change was made. `-F` / `--full-files` attaches the complete working-tree
+contents of every staged file alongside the diff and switches to a dedicated
+prompt (`full_files_prompt.md`) that instructs the LLM to use that full
+context to infer intent rather than just describe the mechanical edit.
+
+```sh
+git cai -F                         # attach full contents of all staged files
+git cai -F -f src/foo.py -f src/bar.py   # only these files
+git cai -f docs/README.md          # restrict diff to one file (no full files)
+```
+
+Both flags log the affected files at INFO level as paths relative to the
+repository root, so you can see exactly what ends up in the prompt — binaries,
+deleted files, and paths matched by `.caiignore` are filtered out and skipped.
+
+Persist the default:
+
+```sh
+git cai -S full_files=true
+```
+
 ### Changing configuration from the CLI
 
 Instead of editing YAML files manually, use `--set` or `--set-home` to update config values.
@@ -259,7 +296,7 @@ Instead of editing YAML files manually, use `--set` or `--set-home` to update co
 
 ```sh
 git cai -S default=anthropic           # change the default provider
-git cai -S emoji=false                  # disable emojis
+git cai -S emoji=false                 # disable emojis
 git cai -S groq.model=llama-3.3-70b    # nested key (dot notation)
 git cai -S openai.temperature=0.7      # set temperature as float
 ```

{git_cai_cli-0.11.2 → git_cai_cli-0.12.1}/cai_config.yml

@@ -1,4 +1,4 @@
-default: anthropic
+default: openai
 language: en
 style: professional
 emoji: true
@@ -7,11 +7,15 @@ branch_context: false
 load_tokens_from: /home/thorsten/.config/cai/tokens.yml
 prompt_file: ""
 squash_prompt_file: ""
+full_files_prompt_file: ""
 token_logging: true
 measure_time: true
+timeout: 30
+full_files: false
 anthropic:
-  model: claude-opus-4-6
+  model: claude-haiku-4-5
   temperature: 0
+  max_tokens: 32768
 deepseek:
   model: deepseek-chat
   temperature: 0
@@ -27,8 +31,9 @@ mistral:
 ollama:
   model: llama3.1
   temperature: 0
+  timeout: 300
 openai:
-  model: gpt-5.2
+  model: gpt-5.4
   temperature: 0
 xai:
   model: grok-4-1-fast-reasoning

{git_cai_cli-0.11.2 → git_cai_cli-0.12.1}/docs/git-cai.txt

@@ -11,7 +11,9 @@ SYNOPSIS
 --------
 [verse]
 `git cai` [-A | --amend] [-a | --all] [-C | --conventional]
-        [-b | --branch] [-c | --crazy] [-d | --debug] [-g | --generate-config]
+        [-b | --branch] [-c | --crazy] [-d | --debug]
+        [-F | --full-files] [-f PATH | --files PATH]
+        [-g | --generate-config]
         [-H KEY=VALUE | --set-home KEY=VALUE] [-h | --help]
         [-i | --install-completion]
         [-l [config|editor|language|model|path|provider|style] |
@@ -19,7 +21,9 @@ SYNOPSIS
         [-m MODEL | --model MODEL]
         [-P PROVIDER | --provider PROVIDER] [-p | --generate-prompts]
         [-S KEY=VALUE | --set KEY=VALUE]
-        [-s [N|HASH] | --squash [N|HASH]] [-t | --time] [-u | --update] [-v | --version]
+        [-s [N|HASH] | --squash [N|HASH]]
+        [-T SECONDS | --timeout SECONDS]
+        [-t | --time] [-u | --update] [-v | --version]
         [-x CONTEXT | --context CONTEXT]
 
 
@@ -154,6 +158,51 @@ Cannot be combined with `--help` or `--version`.
 git cai -d
 ----
 
+-F, --full-files::
+Send the full working-tree contents of every staged file alongside the diff.
+By default, git-cai sends only the staged diff to the LLM. With `-F`, the
+diff is followed by a block of the form:
++
+----
+--- File: path/to/file.py ---
+<full contents>
+----
++
+for each staged text file. This produces higher-quality commit messages for
+large or context-dependent changes, at the cost of more input tokens. Binary
+files and files matching `.caiignore` are skipped.
++
+The flag applies for the current invocation only. To enable it permanently,
+set `full_files: true` in `cai_config.yml` or run:
++
+----
+git cai -S full_files=true
+----
++
+Can be combined with `-f` to scope the content dump to specific files.
++
+----
+git cai -F
+git cai -F -f src/foo.py
+----
+
+-f, --files PATH::
+Restrict the diff (and full-file content, if `-F` is on) to the given path.
+Repeat the flag to pass multiple paths. `.caiignore` patterns still apply on
+top of the whitelist.
++
+Useful when a working tree contains unrelated staged changes and the commit
+should focus on a subset of them, or when you want the LLM to concentrate
+on a specific file.
++
+Cannot be used with `--list`, `--update`, or `--squash`.
++
+----
+git cai -f src/foo.py
+git cai -f src/foo.py -f tests/test_foo.py
+git cai -F -f src/foo.py
+----
+
 -g, --generate-config::
 Write the default `cai_config.yml` into the current directory. This is a
 convenient way to create a repository-specific configuration that overrides
@@ -294,6 +343,32 @@ git cai -s 3
 git cai -s a1b2c3d
 ----
 
+-T, --timeout SECONDS::
+Override the HTTP timeout for the LLM call in this invocation. The default
+is 30 seconds for remote providers and 300 seconds for Ollama (since local
+generation can be significantly slower). The override applies uniformly to
+whichever provider is selected for this run.
++
+To change the default permanently, set `timeout: <seconds>` in
+`cai_config.yml` or the home config:
++
+----
+git cai -H timeout=60
+git cai -S timeout=45
+----
++
+Ollama keeps its own timeout under `ollama.timeout` (default 300). Override
+it separately:
++
+----
+git cai -H ollama.timeout=600
+----
++
+----
+git cai -T 60
+git cai -T 120 -P anthropic
+----
+
 -t, --time::
 Measure and log the wall-clock time taken to generate the commit message.
 The duration is printed to stderr after the LLM responds. Useful for
@@ -543,8 +618,15 @@ Available configuration keys:
 - `squash_prompt_file` -- path to the squash prompt Markdown file
 - `token_logging` -- log token usage after each LLM call (`true`/`false`)
 - `measure_time` -- log generation time (`true`/`false`)
+- `timeout` -- HTTP timeout in seconds for remote LLM calls (default `30`)
+- `full_files` -- send full working-tree contents of staged files alongside
+  the diff (`true`/`false`, default `false`)
 - `<provider>.model` -- model name for a specific provider
 - `<provider>.temperature` -- temperature for a specific provider
+- `anthropic.max_tokens` -- upper bound on Anthropic response tokens
+  (default `32768`)
+- `ollama.timeout` -- HTTP timeout in seconds for Ollama generation calls
+  (default `300`; overrides the global `timeout` for this provider only)
 
 FILES
 -----
@@ -706,6 +788,27 @@ Measure generation time:
 git cai -t
 ----
 
+Send full file contents in addition to the diff:
+
+----
+git cai -F
+git cai -F -f src/foo.py
+----
+
+Limit generation to specific files:
+
+----
+git cai -f src/foo.py
+git cai -f src/foo.py -f tests/test_foo.py
+----
+
+Override the HTTP timeout for a single run:
+
+----
+git cai -T 60
+git cai -T 120 -P anthropic
+----
+
 Install shell completion:
 
 ----