openads-ai 0.3.0 β†’ 0.4.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/README.md CHANGED
@@ -36,7 +36,49 @@ OpenAds is an **open-source CLI tool** that turns any AI model into a marketing
36
36
  | πŸ”Œ **Direct platform access** | Connect your Google Ads, Google Analytics (GA4), and Meta accounts. The AI reads your live data β€” no more copy-pasting reports. |
37
37
  | πŸ€– **Bring your own model** | Use Google Gemini, OpenAI, Claude, or a local model running on your machine. Your choice. |
38
38
  | πŸ›‘οΈ **Nothing goes live without you** | The AI can read freely, but every write operation (campaign change, budget edit) requires your explicit approval. |
39
- | ⚑ **Autonomous loops** | Let the AI research competitors, test ad variants, and generate hypotheses overnight. Review in the morning. |
39
+ | ⚑ **Three experience tiers** | Express for local models, Standard for mid-tier, Full for frontier β€” the same beautiful UI adapts its depth to your model's capability. |
40
+
41
+ ---
42
+
43
+ ## ⚑ Three Experience Tiers β€” One Beautiful Interface
44
+
45
+ OpenAds automatically adapts its entire experience based on your AI model's capability. Same menus, same polish, fundamentally different depth.
46
+
47
+ ```
48
+ β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
49
+ β”‚ ⚑ Express πŸ“Š Standard πŸš€ Full β”‚
50
+ β”‚ Llama 8B Gemini Flash GPT-4o β”‚
51
+ β”‚ Mistral 7B GPT-4o Mini Claude Sonnet β”‚
52
+ β”‚ Phi-3 Llama 70B Gemini Pro β”‚
53
+ β”œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€
54
+ β”‚ System prompt: System prompt: System prompt: β”‚
55
+ β”‚ ~150 words ~300 words ~500 words β”‚
56
+ β”‚ β”‚
57
+ β”‚ Skills: compact Skills: full Skills: full β”‚
58
+ β”‚ (~100 words each) (~400 words each) + advanced β”‚
59
+ β”‚ β”‚
60
+ β”‚ MCP tools: none MCP tools: 6 MCP tools: 11 β”‚
61
+ β”‚ (context-based) (read-only) (read + write) β”‚
62
+ β”‚ β”‚
63
+ β”‚ Menu: 7 items Menu: 9 items Menu: 10 items β”‚
64
+ β”‚ AR: 4 commands AR: 8 commands AR: 12 commands β”‚
65
+ β”‚ β”‚
66
+ β”‚ Output: bullets Output: tables Output: deep β”‚
67
+ β”‚ & quick tables & structured analysis with β”‚
68
+ β”‚ reports cross-platform β”‚
69
+ β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
70
+ ```
71
+
72
+ ### Why tiers matter
73
+
74
+ A marketing manager running **Llama 3.1 8B** locally shouldn't get a *worse* experience β€” they should get a *different but equally polished* experience. Express mode delivers:
75
+
76
+ - **Fewer menu items** β†’ less decision fatigue
77
+ - **Compact skill files** β†’ the model follows instructions reliably
78
+ - **No MCP tool chains** β†’ no multi-step failures
79
+ - **Structured output templates** β†’ consistent, useful results every time
80
+
81
+ When you're ready for more depth, just run `openads` β†’ Settings β†’ Change experience tier.
40
82
 
41
83
  ---
42
84
 
@@ -78,8 +120,10 @@ Paste this command into your Terminal and press **Enter**:
78
120
  ```bash
79
121
  openads setup
80
122
  ```
81
- This launches a beautiful, step-by-step interactive setup wizard where you can:
82
- * **Select your favorite AI model** (Google Gemini, OpenAI, Claude, or a free local model)
123
+ This launches a 6-step interactive setup wizard where you can:
124
+ * **Select your AI model** (Google Gemini, OpenAI, Claude, or a free local model)
125
+ * **Choose your experience tier** (Express ⚑, Standard πŸ“Š, or Full πŸš€ β€” auto-recommended based on your model)
126
+ * **Set your operational mode** (Audit for read-only safety, or Launch for active campaign changes)
83
127
  * **Link your ad accounts** (Google Ads, GA4, and/or Meta Ads)
84
128
  * **Describe your business** (so the AI writes copy tailored exactly to your brand)
85
129
 
@@ -102,6 +146,24 @@ openads setup
102
146
  openads
103
147
  ```
104
148
 
149
+ ### 🏠 Want to run 100% locally and offline?
150
+
151
+ ```bash
152
+ # 1. Install Ollama (free, open-source)
153
+ brew install ollama # Mac
154
+ # or: curl -fsSL https://ollama.com/install.sh | sh # Linux
155
+
156
+ # 2. Pull a model
157
+ ollama pull llama3.1
158
+
159
+ # 3. Install OpenAds and pick "Local AI" + "Express" in setup
160
+ npm install -g openads-ai
161
+ openads setup
162
+
163
+ # 4. Run β€” no internet required
164
+ openads
165
+ ```
166
+
105
167
  ---
106
168
 
107
169
  ## πŸ’‘ What can I do with it?
@@ -142,6 +204,22 @@ Here are some real examples β€” just type what you need:
142
204
 
143
205
  ---
144
206
 
207
+ ## πŸ”¬ Autoresearch β€” Autonomous Marketing Loops
208
+
209
+ Autoresearch is OpenAds' marquee feature: an autonomous loop that generates, scores, and iterates on marketing hypotheses without you lifting a finger.
210
+
211
+ The menu adapts to your tier:
212
+
213
+ | Tier | Available Commands | What you get |
214
+ |---|---|---|
215
+ | ⚑ Express | Generate, Plan, Debug, Fix | 4 commands, flat menu, 2 loop cycles |
216
+ | πŸ“Š Standard | + Discover, Analyze, Ship | 8 commands, 5 phases, 3 loop cycles |
217
+ | πŸš€ Full | + Validate (Predict, Probe, Reason, Scenario) | 12 commands, 6 phases, 5 loop cycles |
218
+
219
+ **Full tier exclusive**: The Validate phase includes multi-persona debates (5 expert personas evaluate your idea), stress-tests (8 personas attack your brief), and adversarial reasoning on key strategy calls.
220
+
221
+ ---
222
+
145
223
  ## 🧠 Memory β€” Gets Smarter Every Session
146
224
 
147
225
  OpenAds remembers what it learns about your business. After each conversation, the AI appends key insights to a plain markdown file at `~/.openads/context/my-business.md`:
@@ -197,20 +275,11 @@ To align with our **Radical Simplicity & Safety by Default** principles, OpenAds
197
275
  * **Audit Mode (Safe / Read-only β€” Recommended)**: Strict read-only guidelines. The AI can pull live campaign statistics, analyze keywords, find budget waste, and recommend landing page optimizations, but is entirely blocked from making active changes.
198
276
  * **Launch Mode (Read-Write)**: Active campaign changes. Authorizes write modifications (e.g. pausing campaigns, scaling bids, altering budgets, or testing ad creative) but strictly mandates a clear visual preview card and an explicit **confirmation prompt (Y/N)** before executing.
199
277
 
200
- Configure your default mode anytime by running `openads setup`.
278
+ Configure your default mode anytime by running `openads` β†’ Settings β†’ Change operational mode.
201
279
 
202
280
  ---
203
281
 
204
- ## πŸ”’ Security & Privacy
205
-
206
- - **Runs 100% locally.** OpenAds is not a cloud service. Nothing leaves your machine except the API calls you authorize.
207
- - **No telemetry.** We don't track usage, store data, or phone home.
208
- - **Your keys stay on disk.** API keys and tokens are saved to `~/.openads/` on your hard drive. They never touch our servers.
209
- - **Explicit approval for all writes.** The AI previews every campaign change before execution. Nothing goes live without your `Y`.
210
-
211
- ---
212
-
213
- ## 🩺 Troubleshooting
282
+ ## 🩺 Diagnostics
214
283
 
215
284
  Run the built-in diagnostics to check your setup:
216
285
 
@@ -218,24 +287,44 @@ Run the built-in diagnostics to check your setup:
218
287
  openads doctor
219
288
  ```
220
289
 
221
- This verifies your config file, API keys, platform connections (live token checks), and required tools like `uvx`.
290
+ This verifies your config file, API keys, platform connections (live token checks), required tools like `uvx`, and **local AI readiness** (checks if Ollama is running and your model is available).
291
+
292
+ ---
293
+
294
+ ## πŸ”’ Security & Privacy
295
+
296
+ - **Runs 100% locally.** OpenAds is not a cloud service. Nothing leaves your machine except the API calls you authorize.
297
+ - **No telemetry.** We don't track usage, store data, or phone home.
298
+ - **Your keys stay on disk.** API keys and tokens are saved to `~/.openads/` on your hard drive. They never touch our servers.
299
+ - **Explicit approval for all writes.** The AI previews every campaign change before execution. Nothing goes live without your `Y`.
222
300
 
223
301
  ---
224
302
 
225
303
  ## πŸ—ΊοΈ Roadmap
226
304
 
305
+ ### Shipped βœ…
227
306
  - [x] Google Ads & GA4 integration via MCP
228
307
  - [x] Meta Ads integration via MCP
229
308
  - [x] Interactive setup wizard with live token verification
230
- - [x] 12 pre-built skills: Ads, CRO, Copywriting, Analytics, Email, Video, Research, Strategy
231
- - [x] Autonomous research loops
309
+ - [x] 24 pre-built marketing skills across 7 categories
310
+ - [x] 12 autonomous research loop commands (Autoresearch)
232
311
  - [x] Published to npm (`npm install -g openads-ai`)
233
312
  - [x] Memory system β€” AI learns about your business over time
234
313
  - [x] Scheduled automations β€” daily health checks, budget alerts, weekly reports
235
- - [ ] Telegram bot gateway β€” talk to your ads from your phone
314
+ - [x] **Three experience tiers** β€” Express ⚑ / Standard πŸ“Š / Full πŸš€
315
+ - [x] Express skill variants β€” compact 100-word skill files optimized for local models
316
+ - [x] Tier-aware menus, prompts, tool chains, and output depth
317
+ - [x] Local AI diagnostics β€” Ollama reachability, model availability checks
318
+ - [x] Settings quick-switch β€” change tier and mode without re-running setup
319
+
320
+ ### Coming Next πŸ”œ
321
+ - [ ] Telegram/WhatsApp bot gateway β€” talk to your ads from your phone
236
322
  - [ ] LinkedIn Ads integration
237
- - [ ] TikTok Ads integration (leveraging their new [TikTok Ads MCP Server](https://digiday.com/media/tiktok-world-ads-mcp-server/))
238
- - [ ] Pinterest Ads integration (leveraging lessons/patterns from their [MCP Ecosystem](https://medium.com/pinterest-engineering/building-an-mcp-ecosystem-at-pinterest-c3b6b1b9e0f6))
323
+ - [ ] TikTok Ads integration (leveraging their [TikTok Ads MCP Server](https://digiday.com/media/tiktok-world-ads-mcp-server/))
324
+ - [ ] Pinterest Ads integration (leveraging [MCP Ecosystem](https://medium.com/pinterest-engineering/building-an-mcp-ecosystem-at-pinterest-c3b6b1b9e0f6))
325
+ - [ ] Visual report dashboards in the terminal
326
+ - [ ] Headless `--print` mode for CI/CD pipelines
327
+ - [ ] Community-contributed skill packs
239
328
 
240
329
  ---
241
330
 
@@ -252,6 +341,7 @@ Read [CONTRIBUTING.md](CONTRIBUTING.md) to get started.
252
341
  1. **Radical Simplicity** β€” Non-technical marketers must feel at home. No forcing users to learn code, prompt engineering, or API error messages.
253
342
  2. **Marketers First** β€” We design around marketing workflows (audits, copy, analysis), not software concepts.
254
343
  3. **Safety by Default** β€” AI should never spend money or publish campaigns without human approval.
344
+ 4. **Model Parity** β€” A $0 local model should deliver a polished experience, not a broken one. Express mode exists to make this real.
255
345
 
256
346
  ---
257
347