Buffaly how-to library
Use these guides when you want to connect tools, extend Buffaly, operate sessions, or configure model/provider behavior.
Start with one concrete setup: one repository, one Google Workspace account, one trusted phone connection, or one repeated task. After the first setup works, ask Buffaly to verify what it can see and what it can safely do before you authorize write actions.
These pages are organized under /docs/how-tos/.... Each page uses the HTML version as the public source of truth and links to nearby topics for deeper workflows.
Connect tools and access
Connect your first tool
Choose the right first integration before you connect repositories, documents, Codex, or trusted-device access.
Connect GitHub
Use GitHub CLI first, verify repository access, and teach Buffaly how to refer to your repositories.
Connect Google Workspace
Choose desktop or hosted OAuth mode, save refresh-token access, and validate the account connection.
Connect from your phone
Choose trusted-device, hosted/staging, or same-network access and verify you are in the right Buffaly instance.
Set up Codex
Configure Codex as Buffaly's source-editing engine for scoped edits, validation, and committed diffs.
Use Codex in Buffaly
Decide when to use Buffaly directly, when to delegate to Codex subagents, and how authentication paths differ.
Build and extend Buffaly
Explore automation options
Choose the right boundary: memory, prompt action, ProtoScript action, native tool, service, MCP binding, web module, or provider.
Add a prompt skill
Turn repeatable instructions into a discoverable prompt-backed capability.
Add a ProtoScript skill
Create typed Buffaly-native actions that can participate in semantic discovery and tool loading.
Add a C# DLL tool
Use compiled .NET code for deterministic SDK, filesystem, protocol, parsing, or process boundaries.
Use context prompts
Load situational guidance when a task needs coding, operations, safety, or other specialized behavior.
Create a new Buffaly agent
Set up a new agent shape with the right prompts, tools, and source-of-truth expectations.
Manage a master prompt
Update the high-level behavioral prompt safely and keep local personalization separate.
Add a scheduled process
Create a scheduled Buffaly process with a safe trigger model, handler, RunData, prompt context, and verification path.
Use runtime model calls from ProtoScript
Ask a model from ProtoScript through Buffaly runtime providers while keeping context isolated and results explicit.
Operate sessions and local installs
Check this install's capabilities
Ask Buffaly to inventory tools, skills, services, model providers, and configuration-dependent capabilities before planning work.
Work with Plan, Scratch, and Tasks
Use Buffaly's lightweight route, working memory, and durable task artifacts without losing context.
Use session artifacts
Understand files, task documents, generated outputs, and other durable artifacts from a session.
Use the session timeline
Read status updates, tool calls, final answers, and evidence in chronological session history.
Search session history
Find earlier conversations, decisions, and evidence across session records.
Use session semantic search
Find relevant session history by meaning instead of exact text.
Manage sessions in Sessions Web
Use the web surface for session inspection and operational maintenance.
Use settings and feature flags
Inspect and configure behavior gates that change what this install can do.
Configure SQL Server semantic storage
Configure SQL Server-backed semantic database entities and session storage paths.
Configure providers and models
Use OpenAI credentials
Configure API keys or Codex backend credentials and understand how they differ.
Add a model provider
Add a provider module with contracts, credentials, layout, and response handling.
Add or remove provider models
Change provider model catalogs safely, including transport, reasoning, deprecation, and validation details.
Configure the embedding provider
Select and validate the provider used for semantic matching and vector-backed workflows.
Configure the Ollama provider
Configure local Ollama or Ollama Cloud endpoints, model rows, feature settings, and provider validation.