foldrun 20.4.0 → 20.5.0

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.
@@ -68,23 +68,31 @@ keep; foldrun's models (the rest of this page) are what to give anyone else.
68
68
 
69
69
  ## Other providers' models
70
70
 
71
- The same key reaches DeepSeek's, Kimi's, OpenAI's and Gemini's models: name
72
- one, and the gateway sends it to that provider on foldrun's models (billed at
73
- its price, under the same limits).
71
+ The same key reaches OpenAI's, Gemini's, DeepSeek's, Kimi's, Z.ai's (GLM),
72
+ Alibaba's (Qwen), MiniMax's and Xiaomi's (MiMo) models: name one, and the
73
+ gateway sends it to that provider on foldrun's models (billed at its price,
74
+ under the same limits).
74
75
 
75
76
  ```sh
76
77
  claude --model deepseek/deepseek-v4 # or /model deepseek/deepseek-v4 in a session
77
78
  claude --model kimi/kimi-k3
78
79
  claude --model openai/gpt-6 # or gpt-6
79
80
  claude --model gemini/gemini-3-pro # or gemini-3-pro
81
+ claude --model glm-5.3 # Z.ai
82
+ claude --model qwen3.8-max # Alibaba
83
+ claude --model MiniMax-M3 # MiniMax
84
+ claude --model mimo-v2.6-pro # Xiaomi
80
85
  ```
81
86
 
82
87
  - **Choosing a provider.** `provider/model` always works (`anthropic/`,
83
- `deepseek/`, `kimi/`, `openai/`, `gemini/`). A bare name with a known prefix
84
- works too: `deepseek-…`, `kimi-…`, `gpt-…`, `gemini-…`, or `claude-…` for
85
- Anthropic. Anything else goes to Anthropic.
86
- - **How each is reached.** DeepSeek and Kimi speak Anthropic's format and are
87
- sent as they are. OpenAI (Responses API) and Gemini (Chat Completions) are
88
+ `openai/`, `gemini/`, `deepseek/`, `kimi/`, `zai/`, `qwen/`, `minimax/`,
89
+ `mimo/`). A bare name with a known prefix works too: `gpt-…`, `gemini-…`,
90
+ `deepseek-…`, `kimi-…`, `glm-…`, `qwen…`, `minimax-…`, `mimo-…`, or
91
+ `claude-…` for Anthropic. Anything else goes to Anthropic.
92
+ - **How each is reached.** DeepSeek, Kimi, Z.ai, Qwen, MiniMax and MiMo speak
93
+ Anthropic's format and are sent as they are, less what each one's API does
94
+ not document (context editing, Claude's tool search, `service_tier`;
95
+ `thinking` in the words each one takes). OpenAI (Responses API) and Gemini (Chat Completions) are
88
96
  translated both ways, by the same translator runs use.
89
97
  - Tools work on all of them.
90
98
  - Gemini's thought signatures are kept between your calls (in the
package/docs/faq.md CHANGED
@@ -18,7 +18,8 @@ Chat Completions API — once for the whole account in Settings → Models
18
18
  block — and the platform bills you nothing for those tokens. On a hosted
19
19
  platform every account brings its own API key — Claude login tokens are not
20
20
  accepted — or, on foldrun's own platform, chooses **foldrun's models**:
21
- Anthropic, OpenAI, Gemini, DeepSeek or Moonshot Kimi on foldrun's keys, billed in credits
21
+ Anthropic, OpenAI, Gemini, DeepSeek, Moonshot Kimi, Z.ai GLM, Alibaba Qwen, MiniMax or
22
+ Xiaomi MiMo on foldrun's keys, billed in credits
22
23
  at foldrun's price list. See [Providers](providers).
23
24
 
24
25
  **Can agents reach the internet?** Yes — the public internet, as themselves,
package/docs/providers.md CHANGED
@@ -32,7 +32,8 @@ Tokens on your own key are billed by your provider, not by the platform
32
32
  ## foldrun's models
33
33
 
34
34
  On foldrun's own hosted platform an account can run on **foldrun's keys**
35
- instead of its own: Anthropic, OpenAI, Google Gemini, DeepSeek or Moonshot Kimi, one at a
35
+ instead of its own: Anthropic, OpenAI, Google Gemini, DeepSeek, Moonshot Kimi, Z.ai GLM,
36
+ Alibaba Qwen, MiniMax or Xiaomi MiMo, one at a
36
37
  time for the whole account. No provider account is needed; the tokens are
37
38
  billed in credits at **foldrun's price list**
38
39
  ([budgets and billing](budgets-and-billing.md#foldruns-models)).
@@ -60,12 +61,16 @@ billed in credits at **foldrun's price list**
60
61
  (`medium` and `xhigh` run as `high`). Kimi takes `low`, `high` and
61
62
  `max`: `medium` and `xhigh` run as `high`, and it always thinks. OpenAI's
62
63
  GPT-6 models (over its Responses API) take all five; Gemini takes `low`,
63
- `medium`, `high`, so `xhigh` and `max` run as `high`. Where a level is
64
- changed the step's log says so. No provider is sent `temperature` or
64
+ `medium`, `high`, so `xhigh` and `max` run as `high`. MiniMax takes all
65
+ five. Qwen 3.8 runs `max` and `high` as `xhigh` (Alibaba's own mapping).
66
+ Z.ai sends effort as written, and documents nothing about it. MiMo has no
67
+ effort setting, so none is sent. Where a level is changed the step's log
68
+ says so. No provider is sent `temperature` or
65
69
  `top_p` — the reasoning models fix them.
66
70
  - **Where each provider processes the prompt** is its own: DeepSeek in
67
71
  the People's Republic of China (its privacy policy), Moonshot in
68
- Singapore, the others as their own terms say. Choose the provider
72
+ Singapore, Alibaba's Qwen in its Singapore region, the others as their own
73
+ terms say. Choose the provider
69
74
  your data may go to.
70
75
  - **What is refused**, before anything is sent: a model not on the price
71
76
  list; the provider's own paid tools (web search, code execution run by
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "foldrun",
3
- "version": "20.4.0",
3
+ "version": "20.5.0",
4
4
  "description": "Agents are just folders. Write them, check them, run them — on your machine.",
5
5
  "license": "FSL-1.1-ALv2",
6
6
  "type": "module",
package/src/commands.mjs CHANGED
@@ -6709,9 +6709,9 @@ async function accountModelCmd(url, positional, flags) {
6709
6709
  out(await put({ source: "own", ...guard }));
6710
6710
  return 0;
6711
6711
  }
6712
- if (source !== "foldrun") throw new Error("`foldrun account model use foldrun --provider <anthropic|openai|gemini|deepseek|moonshot>` or `foldrun account model use own`");
6712
+ if (source !== "foldrun") throw new Error("`foldrun account model use foldrun --provider <anthropic|openai|gemini|deepseek|moonshot|zai|qwen|minimax|mimo>` or `foldrun account model use own`");
6713
6713
  const supply = typeof flags.provider === "string" ? flags.provider : positional[2];
6714
- if (!supply) throw new Error("which provider? `foldrun account model use foldrun --provider anthropic` (anthropic, openai, gemini, deepseek or moonshot — `foldrun account model prices` lists what each costs)");
6714
+ if (!supply) throw new Error("which provider? `foldrun account model use foldrun --provider anthropic` (anthropic, openai, gemini, deepseek, moonshot, zai, qwen, minimax or mimo — `foldrun account model prices` lists what each costs)");
6715
6715
  const r = await put({ source: "foldrun", supply, ...guard });
6716
6716
  out(r);
6717
6717
  return runnable(r) ? 0 : 1;