@globant/coda-darwin-x64 1.0.0

package/assets/docs/cli-reference.md

@@ -0,0 +1,131 @@
+# Commands & Flags
+
+Reference for the slash commands available in the interactive TUI, plus the shell flags you pass when launching CODA from the terminal. Type `/help` inside CODA at any time to see the current list (extensions may add more).
+
+## Session management
+
+| Command | What it does |
+| --- | --- |
+| `/sessions` | Browse, resume, rename, and delete saved sessions |
+| `/new` | Start a fresh session (old one is saved) |
+| `/exit` (alias `/quit`) | Exit CODA; prints `coda --session-id <id>` so you can resume later |
+| `/clear` | Permanently wipe the current session's message history (the session itself stays) |
+
+## Files and changes
+
+| Command | What it does |
+| --- | --- |
+| `/timeline` (alias `/rewind`) | Open the timeline picker to view and restore checkpoints from this session |
+| `/init` | Analyze the project and generate an `AGENTS.md` file |
+| `/project` | Switch the active Glob.AI OS project (Glob.AI OS OAuth only) |
+
+## Configuration
+
+| Command | What it does |
+| --- | --- |
+| `/providers` | Connect, add, update, or remove providers (the setup wizard; same as `--reconfigure`; restart required) |
+| `/switch-profile` (alias `/sp`) | Switch the active provider profile among the ones you've already configured. With no argument it opens the profile picker; `/switch-profile <profile-id>` selects one directly. The change is saved as `activeProfile` and applies on the next launch |
+| `/switch-model` (alias `/sm`) | Open the model picker. `/switch-model <provider/model>` applies for this session; `/switch-model save <provider/model>` also saves it as the default for new sessions |
+| `/login` | Authenticate with Glob.AI OS via browser (OAuth PKCE) when `auth` is configured |
+| `/effort [<level>]` | Set the model's thinking (reasoning) effort — `low`, `medium`, `high`, `xhigh`, `max`. With no level, pick from the levels the current model supports; some models don't support effort at all |
+| `/logout` | Sign out of your Glob.AI OS session |
+| `/settings` | Open the settings panel: UI theme, bash approval, compaction |
+| `/auth-status` (alias `/whoami`) | Show the current authentication status |
+
+## Tools and extensions
+
+| Command | What it does |
+| --- | --- |
+| `/mcp` | Open the MCP manager — add a server, edit `mcp.json`, list servers, view tools, reload |
+| `/mcp status` | Show each server's connection status |
+| `/mcp list-tools [server]` | List the tools a server exposes |
+| `/mcp enable <server>` / `/mcp disable <server>` | Toggle a server for this session |
+| `/skills` | Open the skills manager (no args), or use a subcommand |
+| `/skills list` | List skills and their enabled state |
+| `/skills enable <name>` / `/skills disable <name>` | Toggle a skill |
+| `/skills add <url> [project\|global]` | Install a skill from GitHub |
+| `/skills refresh` | Re-scan the skill directories |
+| `/extensions` | Manage loaded extensions (`list`, `guide`) |
+| `/plugin` | Install, enable, disable, and list plugins |
+| `/reload-plugins` | Reload plugins after you change them on disk |
+| `/agents` | Open the Agents overlay — browse definitions and active runs |
+| `/agents overview` (or `/agents runs`) | Show recent runs and the agent library in chat |
+| `/agents library` | List reusable agent definitions only |
+| `/agents status` | Print whether agents are currently enabled |
+| `/agents enable [global|project]` | Turn agents on (optionally scoped) |
+| `/agents disable [global|project]` | Turn agents off (optionally scoped) |
+| `/agents run [--foreground] <name> <task>` | Start a named agent with the given task (background by default) |
+| `/agents stop <id>` | Cancel a run (a unique ID prefix is enough) |
+| `/agents open <id>` | Inspect a specific run’s details and recent activity |
+| `/agents transcript <id>` | Switch to the child session’s transcript |
+| `/workflows` | Open the workflow dashboard (live run → phase → agent tree and history) |
+| `/workflows stop [runId]` | Stop a workflow run; omit the id to stop all in-flight runs |
+
+## Context
+
+| Command | What it does |
+| --- | --- |
+| `/compact` | Manually condense conversation history to free up context |
+| `/upgrade` | Check for a newer version and open the update dialog |
+| `/help` | Print the full list of available commands |
+
+## Shell flags (run before opening CODA)
+
+| Flag | What it does |
+| --- | --- |
+| `coda --help` / `-h` | Print CLI usage and exit |
+| `coda --version` / `-v` | Print the installed version |
+| `coda --reconfigure` | Open the provider setup wizard (interactive terminal only) |
+| `coda --lastsession` | Resume the most recent session for this project |
+| `coda --session-id <id>` / `-s` | Resume (or create) a specific session |
+| `coda -p "prompt"` / `--prompt` | Run in batch (headless) mode |
+| `coda --prompt-file <path>` / `-pf` | Read the headless prompt from a file |
+| `coda --output text\|json` | Output format for headless runs (`json` emits structured events) |
+| `coda --auto-approve all\|none` | Headless approval handling (use `all` in CI) |
+| `coda --tools <list>` / `-t` | Limit which tools are available for the run |
+| `coda --timeout <ms>` | Abort a headless run after the given time |
+| `coda --checkpoints[=true\|false]` | Force checkpoints on or off for this run |
+| `coda --bash-security <level>` | Set the bash approval level for this run |
+| `coda --model <name>` / `-m` | Use a specific model for this run |
+| `coda -e <path>` | Load an extra extension file for this run |
+| `coda --mcp-config <path>` | Point at a specific `mcp.json` for this run |
+| `coda --system-prompt <text>` | Replace the built-in base system prompt for this run |
+| `coda --append-system-prompt <text>` | Append extra instructions after the normal system prompt |
+| `coda --cwd <dir>` | Override the working directory for this run |
+| `coda --coda-home <dir>` | Use an alternate config/data home (equivalent to `CODA_HOME` env var) |
+| `coda --profile <id>` | Select a persisted profile for this run (read-only) |
+| `coda --provider <name>` | Ephemeral provider for a headless run (`glob-ai \| openai-compat \| ollama`) |
+| `coda --instance <id>` | Glob.AI OS instance for a headless run (e.g. `saas-europe`) |
+| `coda --base-url <url>` | Base URL for `openai-compat` / `ollama` runs |
+| `coda --api-key <key>` | Ephemeral API key for this run (never persisted) |
+
+## Subcommands (run as `coda <command>`)
+
+Beyond launching a session, `coda` exposes a few standalone subcommands:
+
+| Command | What it does |
+| --- | --- |
+| `coda install` | Add the coda binary directory to `PATH` (macOS/Linux: updates shell rc files; Windows: updates user Path env var) |
+| `coda upgrade` | Download and install the platform-native CODA binary from the release feed |
+| `coda logs` | Open the log viewer (or stream when piped); supports filters and `--follow` |
+| `coda logs export` | Write a redacted support bundle to `~/.coda/exports/` — see [View & Share Logs](#logging) |
+| `coda plugin install <source>` | Install a plugin from npm, a Git URL, or a local path. Flags: `--scope global\|project` (default `global`; `project` also records it in `<project>/.coda/plugins.json`), `--name <folder>` (pick a subfolder in a monorepo repo), `--copy` (copy-based install for local paths). Note: a `plugin-id@marketplace-id` argument is **not** supported |
+| `coda marketplace install <source>` | Register a plugin marketplace catalog. Flags: `--id <id>`, `--display-name <name>`, `--root-path <path>` |
+| `coda checkpoints-demo` | Smoke-test the checkpoint subsystem against a throwaway project |
+| `coda logs-demo` | Self-test the logging pipeline end to end |
+
+## Notes and examples
+
+- **Slash commands work only inside the interactive TUI.** Shell flags and subcommands work from your terminal.
+- **`--auto-approve` defaults to `all` in headless runs** so scripts don't hang; pass `--auto-approve none` to refuse anything that would need approval and abort.
+- **`--checkpoints` only records snapshots in headless mode** — there's no `/timeline` picker to restore from without the TUI.
+
+A typical headless invocation pulling several flags together:
+
+```bash
+coda -p "Run the test suite and summarize failures" \
+     --auto-approve all \
+     --output json \
+     --tools read,glob,grep,bash \
+     --timeout 600000
+```

package/assets/docs/cli-vs-batch.md

@@ -0,0 +1,58 @@
+# Interactive vs Headless
+
+You run CODA from your terminal with the `coda` command, which works in two modes. **Interactive** mode opens a full-screen terminal UI (TUI) for back-and-forth work; **headless** (batch) mode runs one prompt and exits, with no UI. Choosing the right one for the job makes a big difference.
+
+## Which one to use
+
+|   | Interactive (TUI) | Headless (Batch) |
+| --- | --- | --- |
+| **How you run it** | `coda` | `coda -p "prompt"` |
+| **Interface** | Full-screen TUI | No UI — stdout only |
+| **Best for** | Day-to-day development | Scripts and CI |
+| **Interaction** | Multi-turn conversation | Single prompt, then exits |
+| **Approval dialogs** | You approve each command | Defaults to `--auto-approve all`; pass `--auto-approve none` to block any unapproved action |
+| **Checkpoints** | On by default | Disabled by default; enable with `--checkpoints`, but `/timeline` rollback requires the TUI |
+
+## Interactive (TUI)
+
+The interactive TUI is how most people use CODA. Open it in your project, start a conversation, and work iteratively. You can review every change before it happens, ask follow-up questions, and navigate the session history.
+
+```bash
+cd my-project
+coda
+```
+
+Use slash commands to control the session: `/sessions`, `/timeline`, `/settings`, `/compact`, and more. See the [Commands & Flags reference](#cli-reference) for the full list.
+
+## Batch (Headless)
+
+Batch mode is for automation. Pass a prompt, CODA runs it, and exits with a non-zero code if something goes wrong — which makes it easy to use in scripts.
+
+```bash
+coda -p "Run the test suite and report failures" --auto-approve all
+```
+
+Batch runs a **single turn** — one prompt in, one result out — then exits. There's no follow-up conversation, no interactive review, and no `/timeline` to roll back from. Because there's no one to answer an approval prompt, the `ask_user` tool is removed and bash auto-approval defaults to `all`. Use `--output json` to get machine-readable events, and check the exit code in your script.
+
+**Alternative prompt sources** — instead of `-p`, you can also pass a file (`--prompt-file PATH` / `-pf`) or pipe stdin into `coda`. All three are mutually exclusive.
+
+**Key batch flags:**
+
+| Flag | What it does |
+| --- | --- |
+| `--auto-approve none` | Refuse any action that needs approval (fail fast; useful for safety checks) |
+| `--timeout MS` | Abort the run after the given milliseconds and exit non-zero |
+| `--tools LIST` | Limit which tools are available (e.g. `read,glob,grep`) |
+| `--bash-security LVL` | Override the bash auto-approve tier for this run |
+| `-s / --session-id ID` | Resume an existing session in headless mode |
+| `--lastsession` | Resume the most recent session in headless mode |
+| `--checkpoints` | Enable checkpoint creation (no TUI restore, but useful for diff workflows) |
+
+See [Automate with Batch Mode](#guide-automate) for full examples.
+
+## How to choose
+
+- **Reach for interactive** when you want to review changes as they happen, ask follow-ups, or work iteratively on something non-trivial. This is the default for day-to-day coding.
+- **Reach for headless** when the task is well-defined and you don't need to watch it: a CI step, a scripted chore across repos, or an unattended analysis you'll read later.
+
+A good rule of thumb: if you'd want to press **Esc** and redirect partway through, use interactive. If you'd be comfortable walking away, headless is fine.

package/assets/docs/config-json.md

@@ -0,0 +1,314 @@
+# config.json — A Complete Example
+
+This page is a **comprehensive, annotated `config.json`** — every block CODA understands, in one place, so you can see how to change a setting by hand. Most people never edit this file directly: `/settings`, `/providers`, `/effort`, and the setup wizard write it for you. But when you want to script a setup, review what a teammate committed, or change something there's no UI for, this is the reference.
+
+For a field-by-field table of the most common keys, see [Configuration Reference](#config-reference). For *how* settings cascade and where the files live, see [Configuration](#configuration).
+
+## Where this file lives
+
+| File | Scope |
+| --- | --- |
+| `~/.coda/config.json` | Your user-global settings (apply in every project) |
+| `<project>/.coda/config.json` | Project overrides (commit to share with the team) |
+| `<project>/coda.config.ts` | TypeScript config — takes priority over JSON if present |
+
+Settings cascade in priority order: **CLI flags → project config → global config → built-in defaults**. A project file only needs the keys it wants to override; everything else falls back to your global file and then to defaults.
+
+> **Every block below is optional.** A real `config.json` only contains the keys you actually set — most commonly `activeProfile` and a `profiles` map (the wizard writes these). The example here is intentionally exhaustive so you can copy the one block you need.
+
+## The complete example
+
+```jsonc
+{
+  // ── Active provider: activeProfile + profiles ──────────────────────────
+  // This is the ONLY provider config CODA writes to disk. "activeProfile" names
+  // the profile currently in use; "profiles" is the map of switchable targets.
+  // Both are managed by /providers and /switch-profile — you rarely hand-edit
+  // them. Each profile's "provider" is one of: "glob-ai" | "openai-compat" | "ollama".
+  "activeProfile": "geai-oauth",
+  "profiles": {
+    // Glob.AI OS via OAuth (browser login). Tokens live in the OS keyring, so
+    // no key is stored here. "instance" is a preset ("clients" | "corp" |
+    // "saas-europe") or a custom id (a custom id also needs its own "baseUrl").
+    "geai-oauth": {
+      "provider": "glob-ai",
+      "label": "Glob.AI OS (OAuth)",
+      "instance": "saas-europe",
+      "auth": { "method": "oauth" },
+      "model": "anthropic/claude-opus-4-8",
+      // Optional per-tier models for the agents "fast"/"smart"/"deep" shortcuts:
+      "fastModel": "anthropic/claude-haiku-4",
+      "smartModel": "anthropic/claude-sonnet-4",
+      "deepModel": "anthropic/claude-opus-4-8"
+    },
+    // Any OpenAI-compatible HTTP API. Key is resolved from ~/.coda/.secrets via secretRef.
+    "my-compat": {
+      "provider": "openai-compat",
+      "label": "My OpenAI-compat server",
+      "baseUrl": "https://api.example.com/v1",
+      "auth": { "method": "apikey", "secretRef": "MY_API_KEY" },
+      "model": "my-model"
+    },
+    // Local Ollama. Keyless; baseUrl optional (defaults to http://127.0.0.1:11434/v1).
+    "local-ollama": {
+      "provider": "ollama",
+      "label": "Local Ollama",
+      "baseUrl": "http://127.0.0.1:11434/v1",
+      "auth": { "method": "none" },
+      "model": "llama3.1"
+    }
+  },
+
+  // UI theme: "classic" or "modern".
+  "theme": "classic",
+
+  // Max agent-loop steps per turn (each step is one model request that may run
+  // tools). Default 300. Raise for very long autonomous runs; lower to cap cost.
+  "maxSteps": 300,
+
+  // Background update check: true (check + notify), false (off), or "notify".
+  "autoupdate": "notify",
+
+  // Error-recovery posture: "balanced" (default), "conservative", "aggressive".
+  "errorHandling": { "posture": "balanced" },
+
+  // NOTE: The top-level "provider" (string) and "providers" (map) keys are
+  // NOT written here anymore. They are a legacy RUNTIME-ONLY layer that CODA
+  // derives from the active profile at launch and never persists to disk. You
+  // don't add them by hand; the profiles above are the source of truth. See
+  // "The legacy providers map" below if you're maintaining an old config.
+
+  // ── Authentication preferences (Glob.AI OS) ─────────────────────────────────
+  "auth": {
+    // Prefer OAuth access tokens or a static API key.
+    "methodPreference": "access_token",
+    "callbackTimeout": 300,
+    "tokenRefreshThreshold": 300
+  },
+
+  // ── Bash tool approvals ───────────────────────────────────────────────
+  // autoApproveLevel: which risk tier runs WITHOUT asking. "safe" < "low" <
+  // "medium" < "high". Truly destructive commands are never auto-approved.
+  // shellMode: "auto" | "bash" | "powershell" | "wsl" (or an explicit path).
+  "bash": {
+    "autoApproveLevel": "low",
+    "shellMode": "auto"
+  },
+
+  // ── Reasoning (thinking) effort ───────────────────────────────────────
+  // effort: "low" | "medium" | "high" | "xhigh" | "max". Per-model support
+  // varies. Change live with /effort.
+  "reasoning": {
+    "enabled": true,  // Default: false
+    "effort": "medium"
+  },
+
+  // ── Compaction (auto context trimming) ────────────────────────────────
+  "compaction": {
+    "enabled": true,
+    "threshold": 0.75,
+    "retainFraction": 0.3
+  },
+
+  // ── Fallback provider/model on failure ────────────────────────────────
+  "fallback": {
+    "enabled": true,  // Default: false
+    "provider": "geai-apikey",
+    "model": "anthropic/claude-sonnet-4"
+  },
+
+  // ── Agents / sub-agents ───────────────────────────────────────────────
+  "agents": {
+    "enabled": true,
+    "maxConcurrent": 6,
+    // "defaultModel" is LEGACY: kept for older configs. Delegated runs now use
+    // your main session model unless an agent definition sets its own "model".
+    "defaultModel": "anthropic/claude-sonnet-4",
+    // Tier overrides used by the "fast"/"smart"/"deep" agent shortcuts:
+    "fastModel": "anthropic/claude-haiku-4",
+    "smartModel": "anthropic/claude-sonnet-4",
+    "deepModel": "anthropic/claude-opus-4-8",
+    // Bundled or discovered agents to hide from run_agent until re-enabled.
+    "disabledDefinitions": []
+  },
+
+  // ── Built-in tools ────────────────────────────────────────────────────
+  "tools": {
+    "glob": { "enabled": true, "max_results": 100 },
+    "grep": {
+      "enabled": true,
+      "backend": "ripgrep"          // or "fastgrep"
+      // "index_path": "/abs/path/to/.coda/fgr"  // fastgrep only
+    }
+  },
+
+  // ── Vision (examine_images) ───────────────────────────────────────────
+  // A LiteLLM-style model id. Set useMainAgentModel:true to reuse the chat model.
+  "vision": {
+    "model": "openai/gpt-5.4"
+  },
+
+  // ── Web search tool ───────────────────────────────────────────────────
+  "webSearch": {
+    "provider": "brave",            // "exa" | "brave" | "serper"
+    "apiKey": "${BRAVE_API_KEY}"
+  },
+
+  // ── Extra extensions to load ──────────────────────────────────────────
+  // Paths to extension modules (in addition to auto-discovered .coda/extensions).
+  "extensions": ["./.coda/extensions/my-tool.ts"],
+
+  // ── Skills tuning ─────────────────────────────────────────────────────
+  "skills": {
+    "descriptionMaxChars": 200,
+    "warnThreshold": 100
+  },
+
+  // ── Checkpoints (file snapshots / undo) ───────────────────────────────
+  "checkpoints": {
+    "enabled": true,
+    "diskSoftCapMb": 2048,
+    "diskHardCapMb": 5120,
+    "maxWorktreeFiles": 100000
+  },
+
+  // ── Session auto-rename ───────────────────────────────────────────────
+  "session": {
+    "autoRenameEnabled": true,
+    "autoRenameTimeoutMs": 5000,
+    "autoRenameMaxRetries": 3
+  },
+
+  // ── Composer (input area) behavior while CODA is busy ─────────────────
+  "composer": {
+    "whileBusy": { "delivery": "queue" },   // "queue" | "steer"
+    "queueSteer": { "enabled": true, "defaultMode": "later" }  // defaultMode: "next" | "later"
+  },
+
+  // ── MCP servers (inline) ──────────────────────────────────────────────
+  // Optional. You can also declare these in ~/.coda/mcp.json or
+  // <project>/.coda/mcp.json (all tiers are merged). stdio uses command+args;
+  // http uses url. See #tools-reference.
+  "mcp": {
+    "servers": {
+      "github": {
+        "transport": "stdio",
+        "command": "npx",
+        "args": ["-y", "@modelcontextprotocol/server-github"],
+        "env": { "GITHUB_TOKEN": "${GITHUB_TOKEN}" }
+      }
+    }
+  },
+
+  // ── Editor / IDE integration (ACP) ────────────────────────────────────
+  "acp": { "enabled": true },
+
+  // ── Logging ───────────────────────────────────────────────────────────
+  "logging": {
+    "level": "info",
+    "levels": { "core.agent": "debug" },
+    "redact": {
+      "denyKeys": ["x-internal-token"],
+      "keySubstrings": ["secret"],
+      "valuePatterns": ["sk-[A-Za-z0-9]+"]
+    },
+    "rawViewer": { "enabled": false }
+  },
+
+  // ── Telemetry (OpenTelemetry export) ──────────────────────────────────
+  "telemetry": {
+    "enabled": false,
+    "exporter": "otlp",
+    "otlpProtocol": "http",         // "http" | "grpc"
+    "otlpHost": "http://localhost:4318",  // collector; omit to use the build default
+    "otlpUseTls": false,          // true for HTTPS/gRPC TLS
+    "serviceName": "coda",        // custom service name in traces
+    "tracesEnabled": true,
+    "metricsEnabled": true
+  }
+}
+```
+
+> **`.json` vs `.jsonc`.** The comments above are for explanation. CODA's `config.json` is parsed leniently, but if your editor flags comments as invalid JSON, strip them — the keys and values are what matter.
+
+## Secrets: keep keys out of the file
+
+Never paste raw API keys into `config.json`. Store them in `~/.coda/.secrets` (dotenv format). API-key profiles reference a secret by **name** via `auth.secretRef`; other blocks (like `webSearch`) use `${VAR}` interpolation:
+
+```bash
+# ~/.coda/.secrets
+MY_API_KEY=sk-...
+BRAVE_API_KEY=...
+```
+
+```jsonc
+// ~/.coda/config.json
+{
+  "activeProfile": "my-compat",
+  "profiles": {
+    "my-compat": {
+      "provider": "openai-compat",
+      "label": "My server",
+      "baseUrl": "https://api.example.com/v1",
+      "auth": { "method": "apikey", "secretRef": "MY_API_KEY" }
+    }
+  }
+}
+```
+
+The `.secrets` file is never committed and its values are scrubbed from logs before anything is written to disk.
+
+## Common edits, by goal
+
+| You want to… | Edit |
+| --- | --- |
+| Switch the active model | `profiles.<id>.model` (or `/model` live) |
+| Add a provider | A new entry under `profiles` (use `/providers`); API-key ones reference a `secretRef` in `.secrets` |
+| Switch which provider is active | `activeProfile` (or `/switch-profile` / `/sp` live) |
+| Switch OAuth ↔ API key | the profile's `auth` block (`{ "method": "oauth" }` vs `{ "method": "apikey", "secretRef": "VAR" }`) |
+| Let bash run more without asking | `bash.autoApproveLevel` (e.g. `"high"`) — see [Permissions & Approvals](#permissions) |
+| Change thinking effort | `reasoning.effort` (or `/effort` live) |
+| Cap or widen parallel agents | `agents.maxConcurrent` (1–10) |
+| Turn off a noisy bundled agent | add its name to `agents.disabledDefinitions` |
+| Use the faster search engine | `tools.grep.backend: "fastgrep"` |
+| Add an MCP server | `mcp.servers` here, or `~/.coda/mcp.json` — see [Tools Reference](#tools-reference) |
+| Quiet or verbose logs | `logging.level` / `logging.levels` |
+| Disable update checks | `autoupdate: false` |
+
+## Environment variables
+
+A couple of settings are controlled by env vars instead of (or on top of) the file:
+
+| Variable | Default | Description |
+| --- | --- | --- |
+| `LOG_LEVEL` | `info` | Log verbosity. A bare token sets the global floor; `service:level` overrides per service (e.g. `info,core.agent:debug`). Overrides `logging.*`. |
+| `CODA_HOME` | `~/.coda` | Alternative home for config, secrets, sessions, logs, and checkpoints. |
+
+## The legacy `providers` map
+
+Older configs (and the runtime, internally) use a top-level **`provider`** string plus a **`providers`** map keyed by provider *type* (`geai`, `openai-compat`, `ollama`, `openai`, `anthropic`, `google`, `groq`, `openrouter`, `azure`, `vertex`). **CODA no longer writes these to disk** — at launch it derives an equivalent runtime map from your active profile. You don't need to add them by hand.
+
+If you're maintaining a config that still uses the old shape, it continues to work, but migrate to `activeProfile` + `profiles` (run `/providers`) when you can. A legacy entry looks like:
+
+```jsonc
+{
+  "provider": "geai-apikey",
+  "model": "anthropic/claude-opus-4-8",
+  "providers": {
+    "geai-apikey": {
+      "type": "geai",
+      "baseUrl": "https://api.geai.globant.com",
+      "apiKey": "${GEAI_API_KEY}"
+    }
+  }
+}
+```
+
+Each entry accepts per-tier model overrides (`fastModel`, `smartModel`, `deepModel`) and a `models` map for capability overrides. See [Configuration Reference](#config-reference) › **providers (runtime client map)** for the full per-type field list.
+
+## See also
+
+- [Configuration](#configuration) — how config works from the UI and how layers cascade.
+- [Configuration Reference](#config-reference) — the field-by-field table.
+- [Connect a Provider](#connect-provider) — setting up providers through the wizard.
+- [Permissions & Approvals](#permissions) — the bash approval levels in detail.