promptforest 0.5.0__py3-none-any.whl → 0.5.0a0__py3-none-any.whl

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.
Files changed (7) hide show
  1. {promptforest-0.5.0.dist-info → promptforest-0.5.0a0.dist-info}/METADATA +15 -12
  2. {promptforest-0.5.0.dist-info → promptforest-0.5.0a0.dist-info}/RECORD +7 -7
  3. {promptforest-0.5.0.dist-info → promptforest-0.5.0a0.dist-info}/WHEEL +0 -0
  4. {promptforest-0.5.0.dist-info → promptforest-0.5.0a0.dist-info}/entry_points.txt +0 -0
  5. {promptforest-0.5.0.dist-info → promptforest-0.5.0a0.dist-info}/licenses/LICENSE.txt +0 -0
  6. {promptforest-0.5.0.dist-info → promptforest-0.5.0a0.dist-info}/licenses/NOTICE.md +0 -0
  7. {promptforest-0.5.0.dist-info → promptforest-0.5.0a0.dist-info}/top_level.txt +0 -0
{promptforest-0.5.0.dist-info → promptforest-0.5.0a0.dist-info}/METADATA RENAMED
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.4
2
2
  Name: promptforest
3
- Version: 0.5.0
3
+ Version: 0.5.0a0
4
4
  Summary: Ensemble Prompt Injection Detection
5
5
  Requires-Python: >=3.8
6
6
  Description-Content-Type: text/markdown
@@ -37,10 +37,6 @@ This discrepancy score enables downstream workflows such as:
37
37
  - Continuous monitoring and model improvement
38
38
 
39
39
  ## Statistics
40
- **E2E Request Latency** \
41
- Average Case: 100ms \
42
- Worst Case: 200ms
43
-
44
40
  PromptForest was evaluated against the SOTA model Qualifire Sentinel model (v2).
45
41
 
46
42
  | Metric | PromptForest | Sentinel v2 |
@@ -60,19 +56,26 @@ PromptForest was evaluated against the SOTA model Qualifire Sentinel model (v2).
60
56
  - Interpretability: Confidence scores can be used to flag uncertain predictions for human review.
61
57
 
62
58
  Interpretation:
63
- While Sentinel has higher raw accuracy, PromptForest provides better-calibrated confidence. For systems where overconfidence on wrong answers is risky, PromptForest can reduce the chance of critical errors despite being smaller and faster.
64
-
65
- Using Sentinel v2 as a baseline, and given that other models perform worse than Sentinel in published benchmarks, PromptForest is expected to offer more reliable and calibrated predictions than most alternatives.
59
+ While Sentinel has higher raw accuracy, PromptForest provides better-calibrated confidence. For systems where overconfidence on wrong answers is risky, PromptForest can reduce the chance of critical errors despite being smaller and faster.
66
60
 
67
61
 
68
62
  ## Supported Models
69
63
 
70
64
  | Provider | Model Name |
71
65
  | ------------- | ----------------------------------------- |
72
- | **Meta** | [Llama Prompt Guard 86M](https://huggingface.co/meta-llama/Prompt-Guard-86M) (Built with Llama) |
73
- | **ProtectAI** | [DebertaV3 Prompt Injection Finetune](https://huggingface.co/protectai/deberta-v3-base-prompt-injection-v2) |
74
- | **Vijil** | [Vijil Dome Prompt Injection Detection](https://huggingface.co/vijil/vijil_dome_prompt_injection_detection) |
75
- | **Appleroll** | [PromptForest-XGB](appleroll/promptforest-xgb) |
66
+ | **Meta** | Llama Prompt Guard 86M (Built with Llama) |
67
+ | **ProtectAI** | DebertaV3 Prompt Injection Finetune |
68
+ | **Vijil** | Vijil Dome Prompt Injection Detection |
69
+ | **Appleroll** | PromptForest-XGBoost |
70
+
71
+ ## Performance
72
+ **E2E Request Latency** \
73
+ Average Case: 100ms \
74
+ Worst Case: 200ms
75
+
76
+ **Accuracy** \
77
+ Preliminary results indicate ensemble performance is at least as good as any individual model. Extensive benchmarking is ongoing.
78
+
76
79
 
77
80
  ## Quick Start
78
81
  To use PromptForest, simply install the pip package and serve it at a port of your choice. It should automatically start downloading the default model ensemble.
{promptforest-0.5.0.dist-info → promptforest-0.5.0a0.dist-info}/RECORD RENAMED
@@ -6,10 +6,10 @@ promptforest/lib.py,sha256=WEuEhNNlRQAerLyEIbTHdi15qdXUMuiQOhfsvaftj4M,9254
6
6
  promptforest/llama_guard_86m_downloader.py,sha256=0B2ttwLWHki0yLEoJG3BwyFE1oqJFY0M2mLEtmMWmPk,1720
7
7
  promptforest/server.py,sha256=uF4Yj7yR_2vEx_7nQabGHGGw-6GWnT0iBZx3UPQK634,2905
8
8
  promptforest/xgboost/xgb_model.pkl,sha256=kSG2r-6TGfhNJfzwklLQOSgG2z610Z5BXxtgQdXE8Vk,2116991
9
- promptforest-0.5.0.dist-info/licenses/LICENSE.txt,sha256=GgVl4CdplCpCEssTcrmIRbz52zQc0fdcSETZp34uBF4,11349
10
- promptforest-0.5.0.dist-info/licenses/NOTICE.md,sha256=XGjuV5VAWBinW6Jzu7-9h0Ph3xwCNzcJdbMH_EgU_g4,356
11
- promptforest-0.5.0.dist-info/METADATA,sha256=fEgp4u7q-P74Zo3eF0gnEjVSFMuIc9z9g-1AoAKPAZs,5002
12
- promptforest-0.5.0.dist-info/WHEEL,sha256=wUyA8OaulRlbfwMtmQsvNngGrxQHAvkKcvRmdizlJi0,92
13
- promptforest-0.5.0.dist-info/entry_points.txt,sha256=sVcjABvpA7P2fXca2KMZSYf0PNfDgLt1NHlYFMPO_eE,55
14
- promptforest-0.5.0.dist-info/top_level.txt,sha256=NxasbbadJaf8w9zaRXo5KOdBqNA1oDe-2X7e6zdz3k0,13
15
- promptforest-0.5.0.dist-info/RECORD,,
9
+ promptforest-0.5.0a0.dist-info/licenses/LICENSE.txt,sha256=GgVl4CdplCpCEssTcrmIRbz52zQc0fdcSETZp34uBF4,11349
10
+ promptforest-0.5.0a0.dist-info/licenses/NOTICE.md,sha256=XGjuV5VAWBinW6Jzu7-9h0Ph3xwCNzcJdbMH_EgU_g4,356
11
+ promptforest-0.5.0a0.dist-info/METADATA,sha256=aqXEuIGz910-Zd-wvaOxCQDvlTDtJpDQYqZAcMGz_HE,4729
12
+ promptforest-0.5.0a0.dist-info/WHEEL,sha256=wUyA8OaulRlbfwMtmQsvNngGrxQHAvkKcvRmdizlJi0,92
13
+ promptforest-0.5.0a0.dist-info/entry_points.txt,sha256=sVcjABvpA7P2fXca2KMZSYf0PNfDgLt1NHlYFMPO_eE,55
14
+ promptforest-0.5.0a0.dist-info/top_level.txt,sha256=NxasbbadJaf8w9zaRXo5KOdBqNA1oDe-2X7e6zdz3k0,13
15
+ promptforest-0.5.0a0.dist-info/RECORD,,