feat: tokenizers (#407)

* feat: autobuild tiktoken lib and schenanigans

Signed-off-by: Aaron Pham <contact@aarnphm.xyz>

* chore: revert readme changes

Signed-off-by: Aaron Pham <contact@aarnphm.xyz>

* fix(build): windows

Signed-off-by: Hanchin Hsieh <me@yuchanns.xyz>

* chore(plugin): early load commands and base setup

Signed-off-by: Aaron Pham <contact@aarnphm.xyz>

* fix(build): make sync

Signed-off-by: Aaron Pham <contact@aarnphm.xyz>

* feat: rust go vroom vroom

Signed-off-by: Aaron Pham <contact@aarnphm.xyz>

* feat: scuffed afaf implementation binding go brrrr

Signed-off-by: Aaron Pham <contact@aarnphm.xyz>

* chore: remove dups

Signed-off-by: Aaron Pham <contact@aarnphm.xyz>

* fix(tokens): calculate whether we should do prompt_caching (fixes #416)

Signed-off-by: Aaron Pham <contact@aarnphm.xyz>

* chore: ignore lockfiles

Signed-off-by: Aaron Pham <contact@aarnphm.xyz>

* Update README.md

* Update crates/avante-tokenizers/README.md

* chore: remove unused

Signed-off-by: Aaron Pham <contact@aarnphm.xyz>

* chore: remove auto build

Signed-off-by: Aaron Pham <contact@aarnphm.xyz>

---------

Signed-off-by: Aaron Pham <contact@aarnphm.xyz>
Signed-off-by: Hanchin Hsieh <me@yuchanns.xyz>
Co-authored-by: yuchanns <me@yuchanns.xyz>
This commit is contained in:
Aaron Pham
2024-08-31 07:19:59 -04:00
committed by GitHub
parent 81b44e4533
commit d2095ba267
27 changed files with 3554 additions and 27 deletions

View File

@@ -6,6 +6,7 @@ local P = require("avante.providers")
local M = {}
M.api_key_name = "ANTHROPIC_API_KEY"
M.tokenizer_id = "gpt-4o"
---@param prompt_opts AvantePromptOptions
M.parse_message = function(prompt_opts)
@@ -28,8 +29,10 @@ M.parse_message = function(prompt_opts)
local user_prompt_obj = {
type = "text",
text = user_prompt,
cache_control = { type = "ephemeral" },
}
if Utils.tokens.calculate_tokens(user_prompt_obj.text) > 1024 then
user_prompt_obj.cache_control = { type = "ephemeral" }
end
table.insert(message_content, user_prompt_obj)
end