dootisaha25 commited on
Commit
c25ca54
·
1 Parent(s): 0a78415

Update: Readme & User Manual

Browse files
Files changed (2) hide show
  1. README.md +11 -6
  2. USER_MANUAL.md +2 -3
README.md CHANGED
@@ -18,9 +18,12 @@ license: mit
18
 
19
  ---
20
 
21
- ## Overview
22
 
23
- Startup Survival Simulator is a real-world, OpenEnv-compliant environment exposing a standard `reset()` / `step()` / `state()` interface via FastAPI. An AI agent observes 10 startup metrics and chooses one of 7 actions each turn. The environment evolves through compounding effects on growth, revenue, product quality, morale, and cash — reflecting the real decisions an early-stage founder faces.
 
 
 
24
 
25
  Episodes end when the startup **goes bankrupt**, **reaches 10,000 users**, or hits the **50-step timeout**.
26
 
@@ -53,13 +56,13 @@ export HF_TOKEN="hf_xxxxxxxxxxxx"
53
  | `revenue` | `float` | Revenue this step in USD |
54
  | `growth_rate` | `float [0,1]` | New-user multiplier |
55
  | `burn_rate` | `float` | Operating cost per step in USD |
56
- | `churn_rate` | `float [0,1]` | Fraction of users lost per step |
57
  | `product_quality` | `float [0,1]` | Product quality score |
58
- | `market_demand` | `float [0,1]` | External market demand score |
59
  | `morale` | `float [0,1]` | Team morale score |
60
  | `time_step` | `int` | Current step counter |
61
 
62
- **Starting values:** cash=50,000 · users=100 · revenue=1,000 · growth_rate=0.08 · burn_rate=4,500 · churn_rate=0.03 · product_quality=0.55 · market_demand=0.60 · morale=0.70
 
 
63
 
64
  ---
65
 
@@ -68,11 +71,13 @@ export HF_TOKEN="hf_xxxxxxxxxxxx"
68
  | Action | Effect |
69
  |---|---|
70
  | `increase_marketing` | +growth_rate, +market_demand, ++burn_rate |
71
- | `hire_engineer` | ++product_quality, +morale, +++burn_rate |
72
  | `improve_product` | +product_quality, −churn_rate, +morale |
73
  | `reduce_costs` | −burn_rate, −growth_rate, −morale |
74
  | `pivot_market` | Random market_demand ± shift (high risk/reward) |
75
  | `raise_funding` | Probabilistic +$30,000 cash (based on product quality & users) |
 
 
76
  | `do_nothing` | −morale (tiny) |
77
 
78
  ---
 
18
 
19
  ---
20
 
21
+ Startup Survival Simulator is a real-world, OpenEnv-compliant environment exposing a standard `reset()` / `step()` / `state()` interface via FastAPI. An AI agent observes 8 startup metrics and chooses one of 9 actions each turn.
22
 
23
+ The environment is designed to test:
24
+ - **Partial Observability:** Critical metrics like `market_demand` and `churn_rate` are hidden. The agent must orchestrate multi-step workflows by using the `analyze_market` tool (API) to pierce this fog.
25
+ - **Mistakes & Recovery:** Rapid growth builds hidden `technical_debt`. Without using the `refactor_code` action, the startup will experience a massive "Server Crash".
26
+ - **Sparse Rewards:** The agent receives 0 reward per step, requiring successful scaling to hit massive milestone payouts.
27
 
28
  Episodes end when the startup **goes bankrupt**, **reaches 10,000 users**, or hits the **50-step timeout**.
29
 
 
56
  | `revenue` | `float` | Revenue this step in USD |
57
  | `growth_rate` | `float [0,1]` | New-user multiplier |
58
  | `burn_rate` | `float` | Operating cost per step in USD |
 
59
  | `product_quality` | `float [0,1]` | Product quality score |
 
60
  | `morale` | `float [0,1]` | Team morale score |
61
  | `time_step` | `int` | Current step counter |
62
 
63
+ *Note: `market_demand`, `churn_rate`, and `technical_debt` are explicitly hidden from the state to enforce World Modeling.*
64
+
65
+ **Starting values:** cash=50,000 · users=100 · revenue=1,000 · growth_rate=0.08 · burn_rate=4,500 · product_quality=0.55 · morale=0.70
66
 
67
  ---
68
 
 
71
  | Action | Effect |
72
  |---|---|
73
  | `increase_marketing` | +growth_rate, +market_demand, ++burn_rate |
74
+ | `hire_engineer` | ++product_quality, +morale, +++burn_rate, +tech_debt |
75
  | `improve_product` | +product_quality, −churn_rate, +morale |
76
  | `reduce_costs` | −burn_rate, −growth_rate, −morale |
77
  | `pivot_market` | Random market_demand ± shift (high risk/reward) |
78
  | `raise_funding` | Probabilistic +$30,000 cash (based on product quality & users) |
79
+ | `analyze_market` | Tool action: Cost $1,000. Returns noisy info about hidden market_demand and churn_rate |
80
+ | `refactor_code` | Recovery action: Cost $2,500. Reduces hidden tech_debt to prevent Server Crashes |
81
  | `do_nothing` | −morale (tiny) |
82
 
83
  ---
USER_MANUAL.md CHANGED
@@ -190,9 +190,7 @@ Omitting the body or sending `{}` uses the default seed (42).
190
  "revenue": 1000.0,
191
  "growth_rate": 0.08,
192
  "burn_rate": 4500.0,
193
- "churn_rate": 0.03,
194
  "product_quality": 0.55,
195
- "market_demand": 0.6,
196
  "morale": 0.7,
197
  "time_step": 0
198
  }
@@ -211,7 +209,8 @@ Applies one action and advances the simulation by one time step.
211
  **Valid action values:**
212
  ```
213
  increase_marketing | hire_engineer | improve_product |
214
- reduce_costs | pivot_market | raise_funding | do_nothing
 
215
  ```
216
 
217
  **Response:**
 
190
  "revenue": 1000.0,
191
  "growth_rate": 0.08,
192
  "burn_rate": 4500.0,
 
193
  "product_quality": 0.55,
 
194
  "morale": 0.7,
195
  "time_step": 0
196
  }
 
209
  **Valid action values:**
210
  ```
211
  increase_marketing | hire_engineer | improve_product |
212
+ reduce_costs | pivot_market | raise_funding |
213
+ analyze_market | refactor_code | do_nothing
214
  ```
215
 
216
  **Response:**