telize 0.2.2__tar.gz → 0.3.0__tar.gz

{telize-0.2.2 → telize-0.3.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: telize
-Version: 0.2.2
+Version: 0.3.0
 Summary: Automate with flows, not loose prompts. Chain LLM, shell, and code in one YAML pipeline.
 Project-URL: Homepage, https://github.com/telize-ai/telize
 Project-URL: Documentation, https://github.com/telize-ai/telize#readme
@@ -36,50 +36,50 @@ Requires-Dist: ruff>=0.9; extra == 'dev'
 Requires-Dist: types-pyyaml>=6.0; extra == 'dev'
 Description-Content-Type: text/markdown
 
-# Telize
+# 🚀 Telize
 
 **Build reproducible, structured AI workflows with YAML and run them from your terminal, combining LLMs, shell, Python, and more—fully under your control.**
 
 Telize is a low-code framework for building agent-style pipelines: chain shell commands, file I/O, LLM calls, Python functions, and nested flows in a single workflow file. Configuration is validated before execution, and the CLI shows live progress as each step completes.
 
-![Telize CLI demo](examples/show.gif)
+![Telize CLI demo](https://raw.githubusercontent.com/telize-ai/telize/main/examples/show.gif)
 
 [CI](https://github.com/telize-ai/telize/actions/workflows/ci.yml) · [Python 3.12+](https://www.python.org/downloads/) · [License](LICENSE)
 
 ---
 
-## Table of contents
+## 🧭 Table of contents
 
-- [Features](#features)
-- [Requirements](#requirements)
-- [Installation](#installation)
-- [Quick start](#quick-start)
-- [Motivation](#-motivation)
-- [How it works](#how-it-works)
-- [Workflow reference](#workflow-reference)
-- [Examples](#examples)
-- [CLI](#cli)
-- [Development](#development)
-- [Contributing](#contributing)
-- [License](#license)
+- [✨ Features](#-features)
+- [🧩 Requirements](#-requirements)
+- [📦 Installation](#-installation)
+- [⚡ Quick start](#-quick-start)
+- [🚀 Motivation](#-motivation)
+- [⚙️ How it works](#-how-it-works)
+- [📚 Workflow reference](#-workflow-reference)
+- [🧪 Examples](#-examples)
+- [💻 CLI](#-cli)
+- [🛠️ Development](#-development)
+- [🤝 Contributing](#-contributing)
+- [📄 License](#-license)
 
-## Features
+## ✨ Features
 
 - **YAML workflows** — one file defines `config`, named `models`, flows, and steps
 - **Composable steps** — `input`, `llm`, `shell`, `python`, `flow`, and `yaml` actions
 - **Jinja templating** — wire step outputs together with `{{ steps.name.output }}`
-- **Loops and sub-flows** — iterate LLM steps over split lists; call nested flows with `uses: flow`
+- **Loops and sub-flows** — add `loop` to any step to iterate it over split lists; call nested flows with `uses: flow`
 - **Validated upfront** — Pydantic models catch schema errors before any step runs
 - **Rich CLI output** — progress, step panels, and errors in the terminal
 - **OpenAI-compatible LLMs** — official OpenAI API or local [Ollama](https://ollama.com/) via the same client
 
-## Requirements
+## 🧩 Requirements
 
 - **Python 3.12+**
 - **LLM endpoint** for `uses: llm` steps — [OpenAI](https://platform.openai.com/) or [Ollama](https://ollama.com/); set `api_url` on each model profile (default `http://localhost:11434`)
 - Optional: [uv](https://docs.astral.sh/uv/) for fast local development
 
-## Installation
+## 📦 Installation
 
 ```bash
 pip install telize
@@ -100,7 +100,7 @@ Check the install:
 telize --version
 ```
 
-## Quick start
+## ⚡ Quick start
 
 **1.** For local models, start [Ollama](https://ollama.com/) and pull a model:
 
@@ -163,7 +163,7 @@ By treating LLMs as just another step in a standard automation pipeline, it brin
 
 ---
 
-### ✨ Why it works:
+### 🧠 Why it works
 - **Deterministic Structure + Non-Deterministic AI:**  
   It keeps the overall architecture _rigid and predictable_ (`YAML`), while allowing the AI to handle the _fuzzy, creative tasks_ (**text generation, summarization**) within strict boundaries.
 - **Upfront Validation:**  
@@ -185,7 +185,7 @@ By treating LLMs as just another step in a standard automation pipeline, it brin
 
 ---
 
-### ⚡ Where Telize Might Struggle (_The Limitations_)
+### ⚠️ Where Telize Might Struggle (_The Limitations_)
 While it is great for **structured automation**, it isn’t a silver bullet:
 
 - **Dynamic Decision Making:**  
@@ -195,7 +195,7 @@ While it is great for **structured automation**, it isn’t a silver bullet:
 
 ---
 
-## How it works
+## ⚙️ How it works
 
 1. Telize loads your YAML and validates it against typed Pydantic models.
 2. The flow named in `config.entrypoint` runs first.
@@ -203,9 +203,9 @@ While it is great for **structured automation**, it isn’t a silver bullet:
 4. Later steps can reference earlier outputs via Jinja templates.
 5. The CLI prints progress and results as the workflow runs.
 
-## Workflow reference
+## 📚 Workflow reference
 
-### Top-level structure
+### 🧱 Top-level structure
 
 | Key      | Description                                                         |
 | -------- | ------------------------------------------------------------------- |
@@ -213,13 +213,14 @@ While it is great for **structured automation**, it isn’t a silver bullet:
 | `models` | Named LLM profiles; referenced by `model:` on each `uses: llm` step |
 | `flows`  | Named flows; `config.entrypoint` must match one of these keys       |
 
-### `config`
+### ⚙️ `config`
 
 | Field        | Description                                       |
 | ------------ | ------------------------------------------------- |
 | `entrypoint` | Name of the flow to run when the file is executed |
+| `repeat`     | Optional repeat interval in seconds. Omitted, `null`, or `-1`: run once. `0`: restart immediately after each run finishes. `N > 0`: restart `N` seconds after each run started; if a run exceeds `N` seconds, restart immediately when it finishes |
 
-### `models`
+### 🤖 `models`
 
 Each key under `models` is a profile name (for example `default`, `creative`). LLM steps pick a profile with `model: <name>`.
 
@@ -258,24 +259,32 @@ models:
     api_url: http://{{ env.OLLAMA_HOST }}:11434
 ```
 
-### Flow
+### 🌊 Flow
 
 | Field   | Description                                               |
 | ------- | --------------------------------------------------------- |
 | `steps` | List of steps (unique `name` per flow), executed in order |
 
-### Steps (`uses`)
+Every step also supports:
+
+| Field        | Description                                                                 |
+| ------------ | --------------------------------------------------------------------------- |
+| `name`       | Unique id within the flow; referenced as `{{ steps.<name>.output }}`        |
+| `output_to`  | Optional path (relative to the workflow file); raw step output is written when the step finishes |
+| `loop`       | Optional; run the step once per item (`items` split by `split_by`, default `\n<|separator|>\n`), exposing each as `{{ item }}` and joining outputs with `separator` (default `\n<|separator|>\n`) |
+
+### 🪜 Steps (`uses`)
 
 | `uses`   | Description                                                                                                                               |
 | -------- | ----------------------------------------------------------------------------------------------------------------------------------------- |
-| `input`  | Read a `file` or a `directory` (with glob `include`)                                                                                      |
-| `llm`    | Send a `prompt` using a named `model` from `models`; optional `output_to`, `loop`                                                         |
+| `input`  | Read a `file` or a `directory` (with glob `include`; optional `separator` when joining directory files, default `\n<|separator|>\n`)                   |
+| `llm`    | Send a `prompt` using a named `model` from `models`                                                                                       |
 | `shell`  | Run `run` commands; optional `envs` (supports templates)                                                                                  |
 | `python` | Call `call` (`module.function`) with `args`                                                                                               |
 | `flow`   | Run another flow via `run`                                                                                                                |
 | `yaml`   | Run an external workflow from `file` (own `config`, `models`, and `flows`); optional `input` map passed to the child as `{{ input.key }}` |
 
-### Templating
+### 🧩 Templating
 
 Telize uses [Jinja2](https://jinja.palletsprojects.com/) in step fields.
 
@@ -303,7 +312,7 @@ Example — chain a shell step into an LLM step:
     {{ steps.fetch_data.output }}
 ```
 
-## Examples
+## 🧪 Examples
 
 | File                                                             | What it demonstrates                                     |
 | ---------------------------------------------------------------- | -------------------------------------------------------- |
@@ -312,7 +321,7 @@ Example — chain a shell step into an LLM step:
 | [`examples/shell_to_llm.yaml`](examples/shell_to_llm.yaml)       | Shell → LLM with `{{ steps.*.output }}`                  |
 | [`examples/read_file.yaml`](examples/read_file.yaml)             | `uses: input` — single file                              |
 | [`examples/read_directory.yaml`](examples/read_directory.yaml)   | `uses: input` — directory glob                           |
-| [`examples/llm_save_output.yaml`](examples/llm_save_output.yaml) | `output_to` — persist LLM text to disk                   |
+| [`examples/llm_save_output.yaml`](examples/llm_save_output.yaml) | `output_to` — persist step output to disk                |
 | [`examples/llm_loop.yaml`](examples/llm_loop.yaml)               | `loop` — split output and iterate                        |
 | [`examples/call_subflow.yaml`](examples/call_subflow.yaml)       | `uses: flow` — sub-flow in the same file                 |
 | [`examples/nested_workflow.yaml`](examples/nested_workflow.yaml) | `uses: yaml` — external workflow + `input`               |
@@ -321,7 +330,7 @@ Example — chain a shell step into an LLM step:
 | [`examples/shell_with_env.yaml`](examples/shell_with_env.yaml)   | Shell `envs` and load-time `{{ env.* }}`                 |
 | [`examples/env_config.yaml`](examples/env_config.yaml)           | `{{ env.VAR }}` in the `models` section at load time     |
 
-## CLI
+## 💻 CLI
 
 ```
 usage: telize [-h] [--version] [-f FILE] [--validate-only]
@@ -333,7 +342,7 @@ options:
   --validate-only    parse and validate without running steps
 ```
 
-## Development
+## 🛠️ Development
 
 ```bash
 uv sync
@@ -345,11 +354,11 @@ uv run mypy
 
 See [CONTRIBUTING.md](CONTRIBUTING.md) for pull request guidelines and [CHANGELOG.md](CHANGELOG.md) for release notes.
 
-## Contributing
+## 🤝 Contributing
 
 Contributions are welcome — bug reports, docs, and pull requests. Please read [CONTRIBUTING.md](CONTRIBUTING.md) and open an [issue](https://github.com/telize-ai/telize/issues) before large changes.
 
-## License
+## 📄 License
 
 Apache License 2.0 — see [LICENSE](LICENSE).
 

{telize-0.2.2 → telize-0.3.0}/README.md

@@ -1,47 +1,47 @@
-# Telize
+# 🚀 Telize
 
 **Build reproducible, structured AI workflows with YAML and run them from your terminal, combining LLMs, shell, Python, and more—fully under your control.**
 
 Telize is a low-code framework for building agent-style pipelines: chain shell commands, file I/O, LLM calls, Python functions, and nested flows in a single workflow file. Configuration is validated before execution, and the CLI shows live progress as each step completes.
 
-![Telize CLI demo](examples/show.gif)
+![Telize CLI demo](https://raw.githubusercontent.com/telize-ai/telize/main/examples/show.gif)
 
 [CI](https://github.com/telize-ai/telize/actions/workflows/ci.yml) · [Python 3.12+](https://www.python.org/downloads/) · [License](LICENSE)
 
 ---
 
-## Table of contents
+## 🧭 Table of contents
 
-- [Features](#features)
-- [Requirements](#requirements)
-- [Installation](#installation)
-- [Quick start](#quick-start)
-- [Motivation](#-motivation)
-- [How it works](#how-it-works)
-- [Workflow reference](#workflow-reference)
-- [Examples](#examples)
-- [CLI](#cli)
-- [Development](#development)
-- [Contributing](#contributing)
-- [License](#license)
+- [✨ Features](#-features)
+- [🧩 Requirements](#-requirements)
+- [📦 Installation](#-installation)
+- [⚡ Quick start](#-quick-start)
+- [🚀 Motivation](#-motivation)
+- [⚙️ How it works](#-how-it-works)
+- [📚 Workflow reference](#-workflow-reference)
+- [🧪 Examples](#-examples)
+- [💻 CLI](#-cli)
+- [🛠️ Development](#-development)
+- [🤝 Contributing](#-contributing)
+- [📄 License](#-license)
 
-## Features
+## ✨ Features
 
 - **YAML workflows** — one file defines `config`, named `models`, flows, and steps
 - **Composable steps** — `input`, `llm`, `shell`, `python`, `flow`, and `yaml` actions
 - **Jinja templating** — wire step outputs together with `{{ steps.name.output }}`
-- **Loops and sub-flows** — iterate LLM steps over split lists; call nested flows with `uses: flow`
+- **Loops and sub-flows** — add `loop` to any step to iterate it over split lists; call nested flows with `uses: flow`
 - **Validated upfront** — Pydantic models catch schema errors before any step runs
 - **Rich CLI output** — progress, step panels, and errors in the terminal
 - **OpenAI-compatible LLMs** — official OpenAI API or local [Ollama](https://ollama.com/) via the same client
 
-## Requirements
+## 🧩 Requirements
 
 - **Python 3.12+**
 - **LLM endpoint** for `uses: llm` steps — [OpenAI](https://platform.openai.com/) or [Ollama](https://ollama.com/); set `api_url` on each model profile (default `http://localhost:11434`)
 - Optional: [uv](https://docs.astral.sh/uv/) for fast local development
 
-## Installation
+## 📦 Installation
 
 ```bash
 pip install telize
@@ -62,7 +62,7 @@ Check the install:
 telize --version
 ```
 
-## Quick start
+## ⚡ Quick start
 
 **1.** For local models, start [Ollama](https://ollama.com/) and pull a model:
 
@@ -125,7 +125,7 @@ By treating LLMs as just another step in a standard automation pipeline, it brin
 
 ---
 
-### ✨ Why it works:
+### 🧠 Why it works
 - **Deterministic Structure + Non-Deterministic AI:**  
   It keeps the overall architecture _rigid and predictable_ (`YAML`), while allowing the AI to handle the _fuzzy, creative tasks_ (**text generation, summarization**) within strict boundaries.
 - **Upfront Validation:**  
@@ -147,7 +147,7 @@ By treating LLMs as just another step in a standard automation pipeline, it brin
 
 ---
 
-### ⚡ Where Telize Might Struggle (_The Limitations_)
+### ⚠️ Where Telize Might Struggle (_The Limitations_)
 While it is great for **structured automation**, it isn’t a silver bullet:
 
 - **Dynamic Decision Making:**  
@@ -157,7 +157,7 @@ While it is great for **structured automation**, it isn’t a silver bullet:
 
 ---
 
-## How it works
+## ⚙️ How it works
 
 1. Telize loads your YAML and validates it against typed Pydantic models.
 2. The flow named in `config.entrypoint` runs first.
@@ -165,9 +165,9 @@ While it is great for **structured automation**, it isn’t a silver bullet:
 4. Later steps can reference earlier outputs via Jinja templates.
 5. The CLI prints progress and results as the workflow runs.
 
-## Workflow reference
+## 📚 Workflow reference
 
-### Top-level structure
+### 🧱 Top-level structure
 
 | Key      | Description                                                         |
 | -------- | ------------------------------------------------------------------- |
@@ -175,13 +175,14 @@ While it is great for **structured automation**, it isn’t a silver bullet:
 | `models` | Named LLM profiles; referenced by `model:` on each `uses: llm` step |
 | `flows`  | Named flows; `config.entrypoint` must match one of these keys       |
 
-### `config`
+### ⚙️ `config`
 
 | Field        | Description                                       |
 | ------------ | ------------------------------------------------- |
 | `entrypoint` | Name of the flow to run when the file is executed |
+| `repeat`     | Optional repeat interval in seconds. Omitted, `null`, or `-1`: run once. `0`: restart immediately after each run finishes. `N > 0`: restart `N` seconds after each run started; if a run exceeds `N` seconds, restart immediately when it finishes |
 
-### `models`
+### 🤖 `models`
 
 Each key under `models` is a profile name (for example `default`, `creative`). LLM steps pick a profile with `model: <name>`.
 
@@ -220,24 +221,32 @@ models:
     api_url: http://{{ env.OLLAMA_HOST }}:11434
 ```
 
-### Flow
+### 🌊 Flow
 
 | Field   | Description                                               |
 | ------- | --------------------------------------------------------- |
 | `steps` | List of steps (unique `name` per flow), executed in order |
 
-### Steps (`uses`)
+Every step also supports:
+
+| Field        | Description                                                                 |
+| ------------ | --------------------------------------------------------------------------- |
+| `name`       | Unique id within the flow; referenced as `{{ steps.<name>.output }}`        |
+| `output_to`  | Optional path (relative to the workflow file); raw step output is written when the step finishes |
+| `loop`       | Optional; run the step once per item (`items` split by `split_by`, default `\n<|separator|>\n`), exposing each as `{{ item }}` and joining outputs with `separator` (default `\n<|separator|>\n`) |
+
+### 🪜 Steps (`uses`)
 
 | `uses`   | Description                                                                                                                               |
 | -------- | ----------------------------------------------------------------------------------------------------------------------------------------- |
-| `input`  | Read a `file` or a `directory` (with glob `include`)                                                                                      |
-| `llm`    | Send a `prompt` using a named `model` from `models`; optional `output_to`, `loop`                                                         |
+| `input`  | Read a `file` or a `directory` (with glob `include`; optional `separator` when joining directory files, default `\n<|separator|>\n`)                   |
+| `llm`    | Send a `prompt` using a named `model` from `models`                                                                                       |
 | `shell`  | Run `run` commands; optional `envs` (supports templates)                                                                                  |
 | `python` | Call `call` (`module.function`) with `args`                                                                                               |
 | `flow`   | Run another flow via `run`                                                                                                                |
 | `yaml`   | Run an external workflow from `file` (own `config`, `models`, and `flows`); optional `input` map passed to the child as `{{ input.key }}` |
 
-### Templating
+### 🧩 Templating
 
 Telize uses [Jinja2](https://jinja.palletsprojects.com/) in step fields.
 
@@ -265,7 +274,7 @@ Example — chain a shell step into an LLM step:
     {{ steps.fetch_data.output }}
 ```
 
-## Examples
+## 🧪 Examples
 
 | File                                                             | What it demonstrates                                     |
 | ---------------------------------------------------------------- | -------------------------------------------------------- |
@@ -274,7 +283,7 @@ Example — chain a shell step into an LLM step:
 | [`examples/shell_to_llm.yaml`](examples/shell_to_llm.yaml)       | Shell → LLM with `{{ steps.*.output }}`                  |
 | [`examples/read_file.yaml`](examples/read_file.yaml)             | `uses: input` — single file                              |
 | [`examples/read_directory.yaml`](examples/read_directory.yaml)   | `uses: input` — directory glob                           |
-| [`examples/llm_save_output.yaml`](examples/llm_save_output.yaml) | `output_to` — persist LLM text to disk                   |
+| [`examples/llm_save_output.yaml`](examples/llm_save_output.yaml) | `output_to` — persist step output to disk                |
 | [`examples/llm_loop.yaml`](examples/llm_loop.yaml)               | `loop` — split output and iterate                        |
 | [`examples/call_subflow.yaml`](examples/call_subflow.yaml)       | `uses: flow` — sub-flow in the same file                 |
 | [`examples/nested_workflow.yaml`](examples/nested_workflow.yaml) | `uses: yaml` — external workflow + `input`               |
@@ -283,7 +292,7 @@ Example — chain a shell step into an LLM step:
 | [`examples/shell_with_env.yaml`](examples/shell_with_env.yaml)   | Shell `envs` and load-time `{{ env.* }}`                 |
 | [`examples/env_config.yaml`](examples/env_config.yaml)           | `{{ env.VAR }}` in the `models` section at load time     |
 
-## CLI
+## 💻 CLI
 
 ```
 usage: telize [-h] [--version] [-f FILE] [--validate-only]
@@ -295,7 +304,7 @@ options:
   --validate-only    parse and validate without running steps
 ```
 
-## Development
+## 🛠️ Development
 
 ```bash
 uv sync
@@ -307,11 +316,11 @@ uv run mypy
 
 See [CONTRIBUTING.md](CONTRIBUTING.md) for pull request guidelines and [CHANGELOG.md](CHANGELOG.md) for release notes.
 
-## Contributing
+## 🤝 Contributing
 
 Contributions are welcome — bug reports, docs, and pull requests. Please read [CONTRIBUTING.md](CONTRIBUTING.md) and open an [issue](https://github.com/telize-ai/telize/issues) before large changes.
 
-## License
+## 📄 License
 
 Apache License 2.0 — see [LICENSE](LICENSE).
 

telize-0.3.0/examples/chat_input.yaml

@@ -0,0 +1,26 @@
+# Pause the workflow and ask the user for input with uses: chat.
+config:
+  entrypoint: main
+
+models:
+  default:
+    provider: openai
+    model: gpt-4o-mini
+    api_url: https://api.openai.com/v1
+
+flows:
+  main:
+    steps:
+      - name: user_chat
+        uses: chat
+        message: |
+          What would you like help with today?
+
+      - name: respond
+        uses: llm
+        model: default
+        prompt: |
+          The user said:
+          {{ steps.user_chat.output }}
+
+          Reply briefly and helpfully.

{telize-0.2.2 → telize-0.3.0}/examples/llm_loop.yaml

@@ -28,6 +28,5 @@ flows:
         loop:
           items: "{{ steps.topics.output }}"
           split_by: ","
-          execution: sequential
         prompt: |
           Reply with one short hashtag for the topic "{{ item }}" (no explanation).

{telize-0.2.2 → telize-0.3.0}/examples/llm_save_output.yaml

@@ -1,4 +1,4 @@
-# Write LLM output to disk with output_to (path relative to this file).
+# Write step output to disk with output_to (path relative to this file).
 #
 # Run:
 #   telize -f examples/llm_save_output.yaml

{telize-0.2.2 → telize-0.3.0}/examples/read_directory.yaml

@@ -22,6 +22,7 @@ flows:
         directory:
           path: ./data/notes
           include: "*.md"
+          separator: "\n\n===\n\n"
 
       - name: merge_summary
         uses: llm

{telize-0.2.2 → telize-0.3.0}/examples/spec_reference.yaml

@@ -15,6 +15,10 @@
 config:
   # entrypoint — name of the flow that runs when you execute this file
   entrypoint: release_pipeline
+  # repeat — optional; omit, null, or -1 to run once. 0 restarts immediately after
+  # each run finishes. N>0 restarts N seconds after each run started (or immediately
+  # if the run took longer than N seconds).
+  # repeat: 0
 
 # models — named LLM profiles referenced by `model:` on llm steps
 models:
@@ -63,7 +67,7 @@ flows:
 
           Draft Documents:
           {{ steps.fetch_context_docs.output }}
-        # output_to — optional path; raw LLM text is written when the step finishes
+        # output_to — optional path; raw step output is written when the step finishes
         output_to: ./reports/documentation_gap_analysis.md
 
       - name: generate_marketing_copy
@@ -89,7 +93,6 @@ flows:
         loop:
           items: "{{ steps.get_keywords.output }}"
           split_by: ","
-          execution: sequential   # or parallel
         prompt: |
           Create 2 social hashtags for this topic: {{ item }}
 

{telize-0.2.2 → telize-0.3.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "telize"
-version = "0.2.2"
+version = "0.3.0"
 description = "Automate with flows, not loose prompts. Chain LLM, shell, and code in one YAML pipeline."
 readme = "README.md"
 license = "Apache-2.0"

telize-0.3.0/src/telize/__version__.py

@@ -0,0 +1 @@
+__version__ = "0.3.0"

{telize-0.2.2 → telize-0.3.0}/src/telize/cli.py

@@ -109,6 +109,8 @@ def main(argv: list[str] | None = None) -> None:
             observer=observer,
             workflow_input=workflow_input,
         ).run()
+    except KeyboardInterrupt:
+        sys.exit(130)
     except (ConfigError, ExecutionError) as exc:
         _print_error(str(exc))
         sys.exit(1)

{telize-0.2.2 → telize-0.3.0}/src/telize/config/models/__init__.py

@@ -1,4 +1,5 @@
 from telize.config.models.actions import (
+    ChatStep,
     FlowRefStep,
     InputStep,
     LlmStep,
@@ -13,6 +14,7 @@ from telize.config.models.flow import Flow
 from telize.config.models.spec import WorkflowSpec
 
 __all__ = [
+    "ChatStep",
     "Flow",
     "FlowRefStep",
     "GlobalConfig",

{telize-0.2.2 → telize-0.3.0}/src/telize/config/models/actions.py

@@ -6,7 +6,7 @@ from pydantic import BaseModel, ConfigDict, Field, model_validator
 
 
 class LoopConfig(BaseModel):
-    """Iterate an LLM step over items produced from a prior step's output."""
+    """Iterate a step over items produced from a prior step's output."""
 
     model_config = ConfigDict(extra="forbid")
 
@@ -14,12 +14,12 @@ class LoopConfig(BaseModel):
         description="Jinja template resolving to a delimited list (e.g. `{{ steps.foo.output }}`).",
     )
     split_by: str = Field(
-        default=",",
+        default="\n<|separator|>\n",
         description="Delimiter used to split `items` into separate loop iterations.",
     )
-    execution: Literal["sequential", "parallel"] = Field(
-        default="sequential",
-        description="How loop iterations are scheduled.",
+    separator: str = Field(
+        default="\n<|separator|>\n",
+        description="String inserted between each iteration's output when joining results.",
     )
 
 
@@ -30,6 +30,10 @@ class DirectoryInput(BaseModel):
 
     path: str
     include: str = Field(default="*", description="Glob pattern for files to include.")
+    separator: str = Field(
+        default="\n<|separator|>\n",
+        description="String inserted between each file section when joining.",
+    )
 
 
 class _StepBase(BaseModel):
@@ -41,6 +45,17 @@ class _StepBase(BaseModel):
         min_length=1,
         description="Unique step id within the flow; referenced as `steps.<name>.output`.",
     )
+    output_to: str | None = Field(
+        default=None,
+        description="Optional path to write raw step output when the step completes.",
+    )
+    loop: LoopConfig | None = Field(
+        default=None,
+        description=(
+            "Optional loop config; runs the step once per item, exposing the "
+            "current value as `{{ item }}` and joining outputs with `separator`."
+        ),
+    )
 
 
 class InputStep(_StepBase):
@@ -68,11 +83,6 @@ class LlmStep(_StepBase):
         description="Name of a model defined in the top-level `models` mapping.",
     )
     prompt: str
-    output_to: str | None = Field(
-        default=None,
-        description="Optional path to write raw output after the step completes.",
-    )
-    loop: LoopConfig | None = None
 
 
 class ShellStep(_StepBase):
@@ -106,6 +116,16 @@ class FlowRefStep(_StepBase):
     run: str = Field(description="Name of the flow to execute.")
 
 
+class ChatStep(_StepBase):
+    """Prompt the user for input interactively in the terminal."""
+
+    uses: Literal["chat"] = "chat"
+    message: str = Field(
+        default="",
+        description="Optional message shown before collecting user input (supports Jinja).",
+    )
+
+
 class YamlStep(_StepBase):
     """Run a workflow defined in an external YAML file."""
 
@@ -123,6 +143,6 @@ class YamlStep(_StepBase):
 
 
 Step = Annotated[
-    InputStep | LlmStep | ShellStep | PythonStep | FlowRefStep | YamlStep,
+    InputStep | LlmStep | ShellStep | PythonStep | FlowRefStep | ChatStep | YamlStep,
     Field(discriminator="uses"),
 ]